HK1089639B - Ultrasonic blood vessel measurement apparatus and method - Google Patents

Description

Ultrasonic vascular measurement apparatus and method

Technical Field

The invention relates to a method and a device for processing a digital image of a vascular structure including a blood vessel. And more particularly, to a method for interpreting an ultrasound image of the common carotid artery.

Background

Coronary heart disease (CAD) is narrowing of the arteries that supply blood with oxygen and nutrients to the heart. CAD can cause shortness of breath, angina, and even heart attacks. Narrowing of the artery is usually due to plaque build-up or, in other words, an increase in the degree of atherosclerosis (burden). Plaque buildup can also create a risk of stroke, heart attack, and embolism (embolisms) caused by plaque debris detaching from the wall of the artery and occluding smaller vessels. Shortly after the formation of a blood clot (blood clot), which has not hardened yet and is liable to break, the risk of rupture of the arterial wall and detachment of the part of the clot covering the ruptured fragments is particularly great.

Measurement of the degree of coronary artery atherosclerosis itself is difficult and invasive. Moreover, risk assessment involves measuring the degree of atherosclerosis and its rate of progression. The assessment thus involves a variety of invasive procedures over time. Treatment of CAD also requires an additional invasive procedure to measure the effect of the treatment.

The carotid artery, located in the neck and near the skin, proved to reflect the degree of atherosclerosis of the coronary arteries. Furthermore, studies have shown that reducing the degree of atherosclerosis in coronary arteries will be accompanied by a reduction in the degree of atherosclerosis in the carotid arteries.

One non-invasive method for measuring the extent of atherosclerosis is to analyze ultrasound images of the carotid artery. High resolution B-mode ultrasound scanning can generate such images. Ultrasound images typically provide a multi-slice digital image that includes the carotid artery wall, which is then measured to determine or estimate the degree of atherosclerosis (extend). Other imaging systems may similarly provide digital images of the carotid artery, such as Magnetic Resonance Imaging (MRI) and radio frequency imaging.

The carotid wall includes the intima, which is close to the blood flow and thickens, or appears to thicken, with the deposition of fatty matter and plaque; a medium located near the intima and thickened due to hypertension; and the adventitia, which provides a supporting structure for the arterial wall. The channel in which blood flows is a lumen. The combined thickness of the intima and media layers, or intima-media thickness (IMT), reflects the condition of the artery, accurately identifying or reflecting early stages of atherosclerotic disease.

An ultrasound image typically includes an array of pixels, each pixel having a particular value corresponding to its intensity. The intensity (brightness) of a pixel corresponds to the density of the tissue it represents, with brighter pixels representing denser tissue. Different types of tissue, each with a different density, can thus be distinguished in the ultrasound image. Due to the different densities, the lumen, intima, media and adventitia can be identified in the ultrasound image.

Ultrasound images are typically formed by transmitting sound waves toward the tissue to be measured and measuring the intensity and phase of the sound waves reflected by the tissue. This imaging method has certain limitations and errors. For example, the image may be affected by noise caused by an imperfect sensor. Another cause of error is the attenuation of sound waves reflected from tissue located deep in the body or beneath denser tissue. Random reflections from different objects or tissue edges, particularly those caused by non-planar ultrasound, can also increase noise.

The limitations of ultrasonography complicate the interpretation of ultrasound images. Other systems designed to calculate IMT thickness discard the exact portion of the image while compensating for these limitations. Some IMT measurement systems divide the image into columns (columns) and examine each column, looking for the largest, smallest, or constant portion of the image in order to locate the layers of tissue that comprise the artery wall. Such a system may discard a whole column of image data if the selected wall portion in the column is not readily identifiable. This method cannot utilize other identifiable portions of the arterial wall in the train. Moreover, examining pixel columns individually cannot take advantage of the precise information in adjacent columns from which we can extrapolate, interpolate, or otherwise direct the retrieval of information within a pixel column.

Another limitation of existing methods is that they do not adequately limit the range of pixels retrieved in a column of pixels. Noise and poor image quality may cause all searches of the maximum, minimum, or intensity gradients to produce significantly erroneous results. Limiting the scope of the search is a form of filtering that eliminates results that are not likely to be accurate. Existing methods either do not limit the scope of retrieval of key points (criticalpoints) or employ fixed limits that are not customized, and may even be independent of the image content being analyzed.

What is needed is an IMT measurement method that compensates for limitations in ultrasound imaging methods. It would be an advance in the art to provide an IMT measurement method that compensates for noise and poor image quality while utilizing the precise information in each pixel column. It would be an advance to provide a method for measuring IMT that limits the search range of keypoints to regions where actual tissue or tissue boundaries may be located.

Disclosure of Invention

In view of the foregoing, it is a primary object of embodiments of the present invention to provide a novel method and apparatus for selecting measurements, such as IMT measurements, from ultrasound images of different tissue structures, such as the carotid artery.

It is another object of embodiments of the present invention to reduce errors in measurements by limiting the search of tissue boundaries, such as lumen/intima and media/adventitia boundaries, to regions that may contain them.

It is another object of embodiments of the present invention to limit the search area with a data, or data, that is pre-calculated based on analysis of most measurement areas to improve processing speed and accuracy.

It is another object of embodiments of the present invention to verify the assumed boundary positions with a threshold that reflects the actual composition (make-up) of the image.

It is a further object of embodiments of the present invention to validate putative boundary locations based on their proximity to known features of ultrasound images of tissue structures, such as the carotid artery.

It is another object of embodiments of the present invention to compensate for the slope and tapering of the carotid artery, and the mismatch of the reference image frame (reference) with respect to the axial direction of the artery.

It is another object of embodiments of the present invention to compensate for low contrast and noise by extrapolating and interpolating from high contrast portions of an image to low contrast portions of the image.

It is another object of embodiments of the present invention to determine tissue density information, such as plaque density information, using ultrasound images.

Consistent with the foregoing objects, and in accordance with the invention as embodied and broadly described herein, one embodiment of the present invention discloses an apparatus comprising a computer programmed to run an image processing application and to receive an ultrasound image of a tissue structure, such as an image of a common carotid artery.

The image processing application may perform a process for measuring intima-media thickness (IMT) that provides better measurements, lower demands on user skill, and higher reproducibility. In fact, the intensity varies with the specific tissue composition. However, the maximum difference in intensity is often insufficient to locate the boundary of the anatomical feature. Thus, it has been found that in applying different curve fitting analysis and signal processing techniques, structural boundaries can be clearly defined, even in the face of rather "noisy" data.

In certain embodiments of the methods and apparatus according to the present invention, an ultrasound imaging device or other imaging device, such as a magnetic resonance imaging system (MRI), computed tomography (CT-Scan), radio frequency images, or other mechanism may be used to generate the digital images. Typically, a digital image contains different pixels, each pixel representing a picture element (picture element) at a particular location in the image. Each pixel is recorded with a certain intensity. Typically the intensity ranges between 0 and 255. In alternative embodiments, the pixels may have color and intensity.

In some embodiments, the image is first calibrated in size. That is, the determination of IMT values, e.g., image size, is preferably calibrated for the measurement involved. Thus, the image scale may be used to show a two-dimensional measurement across the image.

In certain embodiments, the ultrasound images are acquired by the patient supine and in a horizontal orientation. Thus, the longitudinal direction of the image is generally horizontal and approximately coincident with the carotid artery axis. The vertical direction of the image corresponds to the approximate direction across the carotid artery.

In some embodiments of the method and apparatus according to the invention, the measurement area may be selected by a user, or by an automated algorithm. A user familiar with the computer imaging appearance of an ultrasound system can quickly select a measurement region. For example, the horizontal center of the image may be selected adjacent to the media/adventitia boundary of the vessel in question.

Lower density materials tend to appear darker in the ultrasound image, absorbing the ultrasound signal from the transmitter and thus providing less return reflection to the sensor. Thus, a user can relatively quickly identify areas of high intensity that represent denser and more reflective material in the outer membrane area, and darker areas that represent low density or lower absorption in the inner lumen area.

In general, a method of characterizing plaque accumulation within a blood vessel may include measurement of apparent intima-media thickness. In one embodiment, the method may include providing an image. The images are typically oriented in a machine direction that extends horizontally relative to the viewer and a cross-machine direction that extends vertically relative to the viewer. The orientation corresponds to an image of the carotid artery taken at the neck of a patient lying on the examination table. Thus, the carotid artery is oriented substantially horizontally. The axial direction is the direction of blood flow in the vessel and the transverse direction is substantially orthogonal thereto. The image typically comprises pixels. Each pixel has a corresponding intensity that is associated with the intensity of the acoustic wave reflected from the object location represented by the selected region of the image produced by the acoustic wave received by the acoustic receiver.

In selected embodiments of the apparatus and method according to the invention, a series of longitudinal positions along the image may be selected and the brightest pixel appearing in the transverse search for each longitudinal position may be identified. The brightest pixel at any longitudinal position is located in the image across the pixel where the image has the highest level of intensity. The brightest pixels may be curve fitted by a curve belonging to the domain in the longitudinal direction, typically including longitudinal points or locations, and having a range of transverse points corresponding to each brightest pixel. Curve fitting of these brightest pixels provides the curve that constitutes the adventitia datum.

The adventitia datum is useful, although it need not be the center, nor the boundary, of the adventitia. However, a polynomial, exponential or other suitable mathematical function may be used to fit the lateral points of the pixel. The curve fitting may also be achieved by a piecewise fitting of the brightest pixel positions distributed along the longitudinal direction. Other curve fits may also be done on the same domain with some other criteria for pixel selection within the curve range. In certain embodiments, a first, second, or third order polynomial may be selected to piecewise curve fit the adventitia data along segments in the longitudinal direction of the image. Other functions may be used for segmentation or other curve fitting of pixels that meet the selected criteria over the domain.

In some embodiments, the lumen data may be located by one of several methods. In one embodiment, the lumen data is found by translating the adventitia data to a lumen in which each pixel along the curve shape has substantially less than some threshold intensity. The threshold may be the lowest intensity of the image. Alternatively, the threshold may be above the lowest intensity of the image pixels, but corresponds to the general regional intensity or limit (bounding limit) therein found in or near the lumen. This lowest intensity of the image can be obtained from a histogram of pixel intensities within the measurement area. In some embodiments, the threshold is set to the intensity of the lowest intensity pixel in the measurement region plus 10% of the difference between the highest and lowest intensities found in the measurement region. In another embodiment, the operator may easily specify the threshold.

In another embodiment, lumen data may be identified by locating pixels that have the lowest intensity near or below certain thresholds. This can be further limited to situations where the next few pixels across have such low intensity also in the lateral (vertical, cross) direction away from the adventitia. By any means, it has been found that the lumen data comprises a curve fit of pixels representing a set of pixels corresponding to some substantially minimum intensity according to a constraint.

In certain embodiments, the media data may be defined or located by fitting another curve to the lateral positions of the media dark pixels, distributed longitudinally, substantially between the lumen data and adventitia data. Finding media dark pixels is evidence of a local minimum intensity between the lumen data and adventitia data for consecutive search pixels in the lateral direction. That is, the image intensity initially tends to increase with distance from the lumen, then it tends to decrease to a local minimum within the medium, and then it again tends to increase as one moves from the medium to the adventitia.

In practice, threshold values for intensity or distance may be used to limit the range of interest for any search or other operation using the image data. For example, a threshold of 10% of the difference between the maximum and minimum intensities in the measurement region has been found to be a suitable minimum threshold to ensure that the media dark pixels found are not actually too close to the lumen. Similarly, the threshold may be set below the maximum intensity in the measurement region to ensure that the smallest ignored arc remains in the non-region of interest near the adventitia when searching for the location of a medium dark pixel. In some cases, adding 25% of the difference between the maximum and minimum intensities to the minimum intensity is a suitable increment for creating the threshold.

In some cases, such as where no suitable local minimum is found, pixels located between the adventitia data and the lumen data or at a midpoint from the adventitia data to the lumen data may be used as locations for media dark pixels. That is, if the actual intensity decreases monotonically from the adventitia to the lumen, then no local minima exist before going to the lumen. In such cases, limiting the media data points considered to be closer to the adventitia than to the midpoint between the lumen data and the adventitia data proves to be an effective filter.

Typically, the media data is a curve fitted to a line of media dark pixels. However, it has been found effective to establish a medium dark pixel temporary curve fit and move all medium dark pixels located between the temporary curve fit and the adventitia datum onto or laterally onto the temporary curve fit. By comparison, those media dark pixels that are displaced from the temporal curve fit into the lumen allow their actual values to be maintained. One physical reason for this filtering concept is the fact that the adventitia boundary is hardly affected by the variations that occur with data noise. Thus, in particular because the actual media/adventitia boundary is of importance, media data for which the weighting would not fit to a point between the provisional curve fit and the adventitia data proved to be an effective filter.

In certain embodiments, the lumen/intima boundary may be determined by locating the maximum local intensity gradient, i.e., determining the maximum rate of change of intensity with respect to the movement or lateral position change from lumen data to media data traversal. It has been found that the point of local steepest ascent in this transverse traverse accurately represents the lumen/intima boundary. A peak (spike) removal operation may be applied to the lumen/intima boundary in order to remove anomalous peaks in the boundary. The resulting boundary may also be curve fit to reduce errors. In certain embodiments, a peak removal operation is performed prior to any curve fitting to improve the accuracy of the final fit.

Similarly, it has been found that the media/adventitia boundary is accurately represented by those points or pixels representing the points of steepest rise in intensity or most rapid change in intensity relative to lateral position in the traversal from the media data to the adventitia data. Obviously, the distance between the lumen/intima boundary and the media/adventitia boundary represents the intima-media thickness. A peak removal operation may be applied to the media/adventitia boundary to eliminate anomalous peaks in the boundary. The final boundary may also be curve fitted to reduce errors. In certain embodiments, the peak removal operation is performed prior to any curve fitting in order to improve the accuracy of the final curve.

In one embodiment of the invention, software techniques are used to automatically determine one or more aspects of a tissue structure, such as the location of atherosclerotic plaques in the carotid artery wall. For example, software techniques may process images, such as ultrasound images of well-known tissue structures, such as the carotid artery, and identify particular anatomical regions therein. Using information about the identified anatomical region, the software techniques can characterize an aspect of the tissue structure with knowledge about the density and location of different features of the tissue structure and/or the characterized aspect. For example, plaque in the carotid artery can be characterized by knowing the typical density and size of plaque and the location of plaque relative to an anatomical model of the carotid artery.

In another embodiment, a software operator specifies a point in a region of interest within a tissue structure, such as a carotid artery wall, and software techniques are used to determine one of the aspects, such as the physical content of atherosclerotic plaques therein.

In another embodiment, a software operator specifies a path or region around an aspect of the tissue structure, and software techniques are used to search inward from the region until the outer boundary of the tissue structure aspect is determined.

In another embodiment, once an aspect of tissue structure, such as plaque, is determined, either automatically, semi-automatically or manually, software algorithms are used to determine density characteristics of the aspect of tissue structure. These may include mean density, peak density, density range (maximum density minus minimum density), total atherosclerotic plaque area, area of the densest region, density histogram, density standard deviation or variance and, but are not limited to, density centroid.

In another embodiment, once the tissue structure aspect density is determined, the density values may be normalized to eliminate the dependency on instrument gain settings when an imaging device, such as an ultrasound device, captures tissue structure images. This can be achieved by normalizing the density values obtained in the following way, 1) the minimum mean intensity of the image (which generally corresponds to the lumen of the carotid artery when the image is a carotid artery image), and/or 2) the maximum mean intensity of the image (which generally corresponds to the adventitia of the distal wall of the carotid artery when the image is a carotid artery image).

In another embodiment, the normalized or non-normalized tissue structure aspect information may be used to build a database of tissue structure aspect density information and events (stroke, heart attack, etc.) for the patient to which the collected density information corresponds to build a risk stratification database (riskstradition database).

In another embodiment, once the tissue structure-wise density is determined, and after an optional normalization process to remove the ultrasound instrument gain setting, the density information may be recorded for comparison with tissue structure-wise densities collected over a large population, which may then be used for risk stratification.

In another embodiment, tissue structure aspect density may be tracked for the patient to determine density velocity as well as aspect formation velocity.

In another embodiment, the tissue structure aspect location may be determined manually or automatically based on known landmarks and recorded for later reference. The location information may also be associated with an organizational structure aspect database to determine the correctness of the organizational structure aspect location.

The foregoing has outlined rather broadly the features and technical advantages of the present invention in order that the detailed description of the invention that follows may be better understood. Additional features and advantages of the invention will be described hereinafter which form the subject of the claims of the invention. It should be appreciated by those skilled in the art that the conception and specific embodiment disclosed may be readily utilized as a basis for modifying or designing other structures for carrying out the same purposes of the present invention. It should also be realized by those skilled in the art that such equivalent constructions do not depart from the spirit and scope of the invention as set forth in the appended claims. The novel features which are believed to be characteristic of the invention, both as to its organization and method of operation, together with further objects and advantages will be better understood from the following description when considered in connection with the accompanying figures. It is to be expressly understood, however, that each of the figures is provided for the purpose of illustration and description only and is not intended as a definition of the limits of the present invention.

Drawings

For a more complete understanding of the present invention, reference is now made to the following descriptions taken in conjunction with the accompanying drawing, in which:

the foregoing and other objects of the present invention will become more apparent from the following description when taken in conjunction with the accompanying drawings. Understanding that these drawings depict only typical embodiments of the invention and are not therefore to be considered to be limiting of its scope, the invention will be described with additional specificity and detail through the use of the accompanying drawings.

FIG. 1 is a schematic diagram of a general purpose computer suitable for use in accordance with the present invention;

FIG. 2 is a schematic diagram of a system suitable for generating and analyzing ultrasound images of the carotid artery;

FIG. 3 is an example of an ultrasound image of the common carotid artery;

FIG. 4 is a simplified representation of certain features of an ultrasound image of the common carotid artery;

FIG. 5 is a schematic block diagram of a computing system and data structure suitable for analyzing ultrasound images in accordance with the present invention;

FIG. 6 is a flow chart of a process suitable for determining certain features in an ultrasound image of an artery in accordance with the present invention;

FIG. 7 is a schematic block diagram of a data structure suitable for executing a preparation module in accordance with the present invention;

FIG. 8 is a schematic representation of a measurement region and a sampling region superimposed on an ultrasound image of an artery wall in accordance with the present invention;

FIG. 9 is a process flow diagram of a process of tuning in accordance with the present invention;

FIG. 10 is a histogram of pixel intensities within a sampling area with labeled threshold locations in accordance with the present invention;

FIG. 11 is a graph of pixel intensity versus their position for a column of pixels in accordance with the present invention;

FIG. 12 is a simplified representation of a portion of a carotid ultrasound image with lumen, media, and adventitia data superimposed thereon, in accordance with the present invention;

FIG. 13 is a process flow diagram of an adventitia locating process in accordance with the present invention;

FIG. 14 is a series of graphs representing pixel intensity versus position for columns of pixels with lines representing processes for compensating for noise and weak contrast, in accordance with the present invention;

FIG. 15 is a flow chart of a lumen localization process according to the present invention;

FIG. 16 is a process flow diagram for a process for compensating for low contrast in accordance with the present invention;

FIG. 17 is a process flow diagram of an alternative lumen localization process in accordance with the present invention;

FIG. 18 is a simplified representation of an ultrasound image of a common carotid artery with a representation of a procedure line using adventitia data to find lumen data (in text rumena), in accordance with the present invention;

FIG. 19 is a process flow diagram of a media data location process in accordance with the present invention;

FIG. 20 is a flow chart representing a process for locating a media dark pixel in a column of pixels in accordance with the present invention;

FIG. 21 is a process flow diagram illustrating an alternative media data location process of the present invention installed;

FIG. 22 is a graphical representation of the process of adjusting the minimum position to find media data;

FIG. 23 is a process flow diagram of a lumen/intimal edge-defining procedure in accordance with the invention;

FIG. 24 is a process flow diagram of a media/adventitia boundary locating process in accordance with the present invention;

FIG. 25 is a schematic block diagram of a data structure suitable for executing a computing module in accordance with the present invention;

FIG. 26 is a process flow diagram of a taper compensation process in accordance with the present invention;

FIG. 27 is a graph showing IMT measurements taken along a measurement region;

FIG. 28 is a graph illustrating the IMT measurement portion used to calculate normalization factors in accordance with the present invention;

FIG. 29 is a graph showing IMT thickness normalization along a portion of a carotid artery, in accordance with the present invention;

FIG. 30 is a block diagram of a system for measuring density information regarding aspects of tissue structure, in accordance with one embodiment of the present invention; and

FIG. 31 is a flow chart illustrating measuring density information regarding aspects of tissue structure, in accordance with one embodiment of the present invention.

Detailed Description

It will be readily understood that the components of the present invention, as generally described and illustrated in the figures herein, could be arranged and designed in a wide variety of different configurations. Thus, the following more detailed description of the embodiments of the present system and method, as represented in FIGS. 1-29, is not intended to limit the scope of the invention, as claimed, but is merely representative of certain presently preferred embodiments in accordance with the invention. These embodiments will be better understood by reference to the drawings wherein like parts are designated by like numerals throughout.

It will be apparent to those skilled in the art that various modifications to the details shown in fig. 1-29 can be readily made without departing from the essential characteristics of the invention. Accordingly, the following description is intended only by way of example, and simply illustrates certain presently preferred embodiments in accordance with the invention as defined by the appended claims.

Referring now to FIG. 1, an apparatus 10 may include a node 11 (client 11, computer 11) that contains a processor 12 or CPU 12. The CPU 12 is operatively connected to a memory device 14. The memory device 14 may include one or more devices such as a hard disk drive 16 or a non-volatile storage device 16, read only memory 18(ROM) and random access (and typically volatile) memory 20 (RAM).

The apparatus 10 may include an input device 22 for receiving input from a user or another device. Similarly, the output means 24 may be provided within the node 11, or may be contactingly provided within the interior of the apparatus 10. A network card 26 (interface card) or port 28 may be provided for connection to an external device, such as a network 30.

Internally, a bus 32 (system bus 32) operatively interconnects the processor 12, the memory device 14, the input device 22, the output device 24, the network card 26, and the port 28. The bus 32 can be regarded as a data carrier. Likewise, the bus 32 may be implemented in a myriad of configurations. For the bus 32 and the network 30, electric wires, optical fiber lines, wireless electromagnetic communication by visible light, infrared light, and radio frequency waves can also be implemented as appropriate.

The input device 22 may include one or more physical embodiments. For example, a keyboard 34 may be used to interact with the user, as may a mouse 36. The touch screen 38, phone 39, or simple phone line 39 may be used for the user or the like to communicate with other devices.

Similarly, the scanner 40 may be used to receive graphical input that may or may not be converted to other character formats. The hard disk drive 41 or other memory device 14 may be used as an input device whether on the network 30 or within a node 11 or some other node 52 (e.g., 52a, 52b, etc.) from another network 50.

The output device 24 may similarly comprise one or more physical hardware units. For example, in general, port 28 may be used to receive input and transmit output from node 11. However, the monitor 42 may provide output to a user for in-process feedback, or to facilitate two-way communication between the processor 12 and the user. A printer 44 or hard drive-46 may be used as the output device 24 for outputting information.

In general, the network 30 to which the node 11 is connected may in turn be connected to another network 50 via a router 48. In general, the two nodes 11, 52 may be on the network 30 or adjacent networks 30, 50, or may be separated by multiple routers 48 and multiple networks 50 as a single node 11, 52 on an internet network (internet). A single node 52 may have different communication capabilities.

In some embodiments, minimal logic capability is available at any node 52. Note that any single node 52, whether tail reference letters, may be referred to as, or all together as, a node 52 or multiple nodes 52.

The network 30 may include one or more servers 54. The server may be used to manage, store, communicate, transfer, access, update, etc., any number of files for the network 30. In general, the server 54 is accessible by all nodes 11, 52 on the network 30. However, other specialized functions, including communications, applications, etc., may be performed by a single server 54 or multiple servers 54.

In general, node 11 needs to communicate with server 54, router 48, or node 52 over network 30. Similarly, the node 11 may need to communicate with certain remote nodes 52 over another network (50) in an internet connection (e.g., the internet). Similarly, the various elements of device 10 may need to communicate data with one another. In general, a communication link may exist between any pair of devices or elements.

By the expression "node" 52, any or all of the nodes 48, 52, 54, 56, 58, 60, 62, 11 may be represented. Thus, any of nodes 52 may include any or all of the elements shown in node 11.

To support distributed processing, or access, directory services nodes 60 may provide directory services, as is known in the art. Thus, directory service node 60 may store the software and data structures necessary to provide directory services to nodes 52 in network 30, and may provide directory services to other nodes 52 in other networks 50.

Directory service node 60 may generally be a server 54 in a network. However, it may be installed at any node 52. To support directory services, the directory service node 52 may generally include a network card 26 for connecting to the network 30, a processor 12 for processing software commands in the executable directory service. Memory means 20 for operational memory and non-volatile storage means 16 such as a hard disk drive 16. Typically, a user is provided with input device 22 and output device 24 to interact with directory services node 60.

Referring to FIG. 1 in one embodiment, a node 11 may be embodied as any digital computer 11, such as a desktop computer 11. The node 11 may communicate with an ultrasound system 62 having a transducer 64, or "soundhead" 64, for transmitting sound waves to the tissue to be imaged and sensing reflected sound waves from the tissue. The ultrasound system 62 then interprets the reflected sound waves to form an image of the tissue. The image is then transmitted to the node 11 for display on the monitor 4 "and/or analysis. Sensor 64 may be positioned adjacent to carotid artery 65, which is located in the neck of patient 66, in order to produce an ultrasound image of the common carotid artery (hereinafter "carotid artery"). Of course, other imaging methods, such as Magnetic Resonance Imaging (MRI), etc., may be used to generate images of the carotid artery 65.

Server 54 may be connected to node 11 via network 30. Server 54 may store the results of the analysis and/or archive other data related to carotid artery measurements and diagnosis of medical conditions.

Fig. 3 is an example of a carotid ultrasound image produced by the ultrasound system 62. The depth of the grey color indicates the reflectivity, i.e. density, of the tissue, the white areas represent the most dense and most reflective tissue, and the black areas represent the least dense or least reflective tissue. The images output by the ultrasound system may also include calibration marks 72a-72e or time stamps (time stamp)72 f.

Referring to fig. 4, an ultrasound image of the carotid artery includes an array of pixels, where each pixel is associated with a numerical value representing the intensity of the pixel (e.g., black, white, gray depth, etc.). Thus, a horizontal direction 74 may be defined as extending along a row or pixel in the image, and a lateral direction 76 may be defined as extending along a similar column. In certain embodiments of the present invention, the transverse direction 76 may be substantially perpendicular to the direction of blood flow in the carotid artery. The horizontal direction 74 may be substantially parallel to the direction of blood flow.

Ultrasound images of the carotid artery generally reveal various essential features of the artery, such as the lumen 78, as well as the intima 80, media 82, and adventitia 84, with the lumen 78 representing the portion of the arterial lumen in which blood flows, all of which form the wall of the artery. The thickness of the intima 80 and media 82 (intima-media thickness or IMT) can be measured to diagnose the risk of the patient for arteriosclerosis, such as coronary artery disease.

The images generally show the artery proximal wall 86 and distal wall 88. The proximal wall 86 is closest to the skin. The distal wall 88 generally provides a clearer image because as the inner membrane 80 and medium 82 are less dense than the outer membrane 84, the acoustic waves reflected from the outer membrane 84 are less disturbing. To image the adjacent wall 86, the sound waves reflected from the inner membrane 80 and medium 82 must pass through the denser outer membrane 84, which interferes with the sound waves to a measurable degree.

As the common carotid artery extends cephalad, it eventually bifurcates into an internal carotid artery and an external carotid artery. Just prior to branching, the common carotid artery has an expansion point 90. The IMT 92 of a segment 94 (the portion of the common carotid artery distal to the heart) about 10 mm long below the inflation point 90 is typically greater than the IMT 96 of a segment 98 (the portion of the common carotid artery proximal to the heart), the segment 98 extending 10 mm to 20 mm from the inflation point 90. This is true, with 88% of the younger population (average age 25 years), the IMT 92 of segment 94 being 14% thicker than the IMT 96 of segment 98. On the other hand, IMT 92 may be as thick as IMT 96 or thinner in 12% of the younger population. In 69% of the elderly population (mean age 55), IMT 92 was 8% thicker than IMT 96. However, in 31% of the elderly population, IMT 92 is as thick as IMT 96 or thinner than IMT 96.

Tapering the IMT away from the branch may introduce uncertainty into the interpretation of the IMT measurement, as changes in the IMT measurement can be attributed simply to the migration of the measurement points.

Moreover, the walls 86, 88 may be angled 100 relative to the horizontal direction 74. Thus, the IMT measurements for the transverse rows of analysis pixels may change due to the orientation of the carotid artery in the image, rather than a change in actual thickness.

Referring to FIG. 5, the memory device 14 coupled to the processor 12 may include an image processing application 110 whose data structures are executable and operable to measure IMT and other data of the carotid artery. The image processing application may include a calibration module 112, an image marking module (preferencingmodule) 114, a preparation module (preparation module)116, a positioning module 118, a calibration module 120, an image quality module 122, and a reporting module 124.

The calibration module 112 may correlate the distance measured on the image with the real world distance. The calibration module 112 typically takes as input the pixel coordinates of two points in the image and the actual distance between the two points. The calibration module then uses these known values to convert other distances measured in the image to true values.

The calibration module 112 may obtain pixel coordinates from the image by looking up the calibration marks 72a-72 e. The actual distance between the markers 72a-72d may be known without user intervention, or the calibration module may prompt the user to enter the distance or retrieve a value from a file or the like. Alternatively, in some embodiments the indicia 72e may represent the distance between the calibration points 72a-72e, such as information on the model of the ultrasound machine 62, zoom mode, etc. may also be displayed, and the calibration module 112 may store calibration factors that map to different ultrasound machines 62 and their different zoom modes, etc.

The calibration module may then calibrate the image based on the known calibration factors for the particular ultrasound machine 62 in the particular zoom mode. The calibration module 112 may also search for "landmarks," such as physical features, patterns, or structures, in an image and perform a calibration based on known distances between landmarks or known landmark dimensions.

The image labeling module 114, preparation module 116, location module 118, and computation module 120 generally interpret the image and obtain IMT measurements, and the like. The operation of these modules will be described in more detail below.

The image quality module 122 may manipulate the image, or selected regions of interest within the image, to remove noise and otherwise improve image quality. For example, the image quality module 122 may apply a low pass filter to remove noise from the image, or use an edge detection or embossing filter (edge highlighting) to highlight edges. In a typical carotid ultrasound image, the layers of tissue extend parallel to the horizontal direction 74. Thus, a transverse filter may be applied in a substantially horizontal direction 74, or a direction parallel to the boundary between tissue layers, to reduce noise in the direction of deflection, thereby preventing loss of edge data indicative of the boundary between different tissue layers.

The image quality module 122 may also notify the user when the image is too noisy to be useful. For example, the image quality module may display a scale, such as a dial indicator, a numerical value, a color-coded indicator, or the like, on the monitor 42 to indicate the quality of the image. In certain embodiments, the image quality module 122 may first assess image quality by locating the portion of the image representing the lumen 78. Because the lumen 78 is filled with blood of substantially constant density, a high quality image of the lumen will have a substantially constant density. Thus, image quality module 122 may calculate and display the standard deviation of the pixel density inside the lumen as an indication of image noise. The smaller the standard deviation of the pixel density, the higher the image quality.

The image quality module 122 may determine the location of the lumen 78 in the same manner as the location module 118, as discussed below. After finding the lumen/intima boundary adjacent to the wall 86 and the distal wall 88, the image quality module 122 may examine the region between the two boundaries to calculate the standard deviation of the lumen pixel intensities. Alternatively, the image quality control module 122 may evaluate a region of a predetermined size having an edge located near the lumen/intima boundary.

Another criterion that the image quality module 122 may use to assess quality is a histogram of pixel intensities in the measurement region, or in other words, the portion of the image where the IMT was measured. Alternatively, a larger area, including the surrounding measurement area, may be used to calculate the histogram. The form of the histogram typically varies according to the image quality. The image quality module 122 may store the image of the histogram generated from the high quality image and display it on the output device 24 along with the histogram of the image being analyzed.

The operator may then be trained to identify "good" histograms in order to determine whether the measurements obtained from a particular image are reliable. The image quality module may also store and display medium quality images and poor quality histograms to assist the operator. Alternatively, the image quality module 122 automatically compares the histogram to stored image high, medium, and/or low quality histograms and identifies their similarities. This may be achieved by pattern matching techniques or the like.

The reporting module 124 may format the results of the calculations and send them to output devices 24, such as the monitor 42, printer 44, hard drive 46, and the like. The image processing application 110 may also store results to a database 126 or retrieve data from a database 126, the database 126 having a database engine 128 for storing, organizing, and retrieving archived data. Database 126, like any of the modules comprising the present invention, may be physically located on the same node 11 or may be located on server 54, or other nodes 52a-52d, and may communicate with node 11 via network 30. The database engine 128 may be any suitable database application known in the art.

The database 126 may store different records 129. The records 129 may include patient records 130. Patient records 130 may store information such as patient age, weight, risk factors, cardiovascular disease, existing IMT measurements, and other relevant medical information. The diagnostic data 131 may provide data to support statistical analysis of the patient's risk of developing cardiovascular disease. For example, diagnostic data 131 may include study results, etc., correlating the likelihood of a patient developing coronary artery disease to IMT measurements and/or other risk factors.

The measurement record 132 may include information about the measurement process itself. For example, the measurement record 132 may include a reference to the ultrasound image being analyzed or the image itself. The measurement log 132 may also include any inputs during the measurement, the name of the operator performing the measurement, the algorithm used to analyze the image, the values of the various parameters used, the date the measurement was taken, the ultrasound machine data, the values of the error sources, and so forth.

The IMT database 133 may archive IMT measurements for use in later interpretation of the ultrasound images. The IMT database 133 may include records 134 of existing measurements, each including an indexed IMT 135. Index IMT 135 may be an IMT value used to characterize record 134. For example, IMT measurements along a portion of the carotid artery may be stored based on the IMT of a normalized point on a single carotid artery. Thus, the index IMT 135 may be the IMT at the normalized point. Alternatively, the average of all IMT measurements of the measurement section may be used as the index IMT 135. IMT measurements 136 may include IMT measurements made at various points along the length of the carotid artery. IMT measurements 136 may be from one ultrasound image, or from an average of multiple ultrasound image measurements. In some embodiments, the IMT measurements 36 may be a polynomial fit of IMT measurements made along portions of the artery.

The memory device 14 may also include other applications 137 and an operating system 138. Operating system 138 may include executable (e.g., programming) and operational (e.g., information) data structures for controlling the various elements that make up node 11. Moreover, it is to be understood that the architectures depicted in FIGS. 2 and 5 are merely exemplary, and that numerous other architectures are possible without departing from the essential characteristics of the present invention. For example, the node 11 may simply be an ultrasound system 62 having at least one memory device 14 and a processor 12. Accordingly, the image processing application 110 and/or the database 126 may be embedded in the ultrasound system 62. Ultrasound 62 may also include a monitor 42, or other graphical display such as an LCD or LED, to provide ultrasound images and computational results.

Referring to fig. 6, the process of locating the essential features of the carotid artery may include the steps shown. It will be understood that the steps and the ordering of the steps included are exemplary only, and that other combinations and orderings of steps are possible without departing from the essential characteristics of the invention.

The process may include an image calibration step 140 to perform the operations described above in conjunction with the calibration module 112. The preliminary step 142 may identify areas representing portions of the image of the adjacent wall 86 or the distal wall 88 to be analyzed. The marking step 143 may calculate a threshold, or other reference value, based on the image used in the later calculations.

The locating process 144 may identify a plurality of tissue layers that form the artery walls 86, 88. It may also locate the boundaries between tissue layers. Thus, the locating process may include an adventitia data locating step 146, which identifies the location of the adventitia 84 and establishes corresponding data. The lumen data positioning step 148 may establish a lumen internal data curve. The media data location step 150 can identify the portion of the arterial wall corresponding to the media and establish corresponding data. Lumen/intima boundary determination step 152 may search for a lumen/intima boundary between the lumen data and the media data. The media/adventitia boundary locating step 154 may search for a media/adventitia boundary between the media data and the adventitia data.

Referring to fig. 7, the modules are executable and programmed to run on the processor 12 and may be stored in the memory device 14. The preparation step 142 may be performed by a preparation module 116, which may include an input module 160, an automation module 162, a reconstruction module 164, and an adaptation module 166.

Referring to fig. 8, the input module 160 may allow a user to select a point 170 in the image that serves as the center of the measurement region 172. The IMT of the pixel columns within the measurement area may be measured and all columns may be averaged together, or combined, to produce a final IMT measurement and other information. Alternatively, the IMT may be fitted with a longitudinal curve. Accordingly, the height 174 of the measurement region 172 may be selected so that it includes at least a portion of the lumen 78, the inner membrane 80, the medium 82, and the outer membrane 84.

The input module 160 may allow a user to specify the width 176 of the measurement region 172. Alternatively, the input module 160 may use only predetermined values. For example, a suitable value is 5 mm, which in most cases is also approximately equal to the diameter of the carotid artery. Whether automatically found, automatically found with limited certainty or found by the user, or specified by the user, the width 176 may also be selected based on image quality. If the image is noisy or the image contrast is poor, a larger width 176 may be used to finally balance the error. In certain embodiments, where width 176 is automatically selected, input module 160 may select the width based on image quality indication data calculated by image quality module 122. Similarly, the operator may be trained to manually adjust the width 176 based on image quality indication data output by the image quality module 122. In some embodiments, the input module may incrementally increase or decrease the width 176 in response to a user input, such as a mouse click or a keyboard stroke.

The input module 160 may also determine which wall to measure by determining the wall of the walls 86, 88 that is closest in lateral distance to the point 170. This can be done by finding the highly identifiable adventitia 84 in each wall 86, 88 and comparing their distance from the point 170. Alternatively, the input module 160 may select the wall 86, 88 of the adventitia 84 that has the highest intensity (or highest average intensity, highest median intensity, etc.) and thus is more likely to have the desired high contrast.

The point 170 may also be used as the center of the sampling region 178. The pixels defined by the sampling region 178 are used to generate a histogram of pixel intensities that other modules use to determine certain thresholds and evaluate image quality. The selected height 180 generally includes the lumen 78 and adventitia 84, as they represent the lowest and highest intensity regions of the image, respectively, and will be relevant to analysis of the histogram. The selected width 182 may provide an appropriate sampling of pixel intensities. In certain embodiments, width 182 is only the same as width 176 of measurement region 172. Suitable values for height 180 have been shown to be between one-half and one-quarter of width 176 of measurement region 172.

The automation and automation module 162 may automatically specify the location of the measurement region 172 and/or the sampling region 178. The automation module 162 may accomplish this in a variety of ways. For example, the automation module may only horizontally center the region 172 in the center of the image. The lateral center of region 172 may be set at the position of the brightest pixel in the lateral column in the center of the image. These brightest pixels will correspond to the adventitia 84 of the wall that has the highest intensity and therefore contrast. Alternatively, for a large number of consecutive pixels, the automation module 162 may locate the lumen 78 by searching for a central column of pixels whose intensity, or average intensity, is below a specified threshold for the intensity of the pixel in the corresponding lumen. One side of the set of pixels may be selected as the center of the measurement region 172 because that side has a high probability of being adjacent to the lumen/intima boundary. The automation module 162 may also adjust the size and location of the measurement region 172 and the sampling region 178 to exclude markers 72a-72f that may be found in the image. The automation module 162 may also adjust the user-selected measurement region 172 and sampling region 178 to avoid such markings 72a-72 f.

Referring again to FIG. 7, the reconstruction module 164 may store all of the specified data in the database 126 regarding user input, such as the location of the point 170 or the size of the region 172 by the user. The reconstruction module 164 may also store signatures that uniquely identify the measured images or store the images themselves. The reconstruction module 164 may also store other input data for algorithms for locating tissue layers or methods for eliminating noise.

The reconstruction module 164 may store this information in any accessible storage location, such as in the database 126 as the measurement data 132, or in the hard drive 46 of the node 11. The reconstruction module 164 may then recover this information and use it to recreate the IMT measurements and their build process. The reconstruction module 164 may also allow the user to adjust the input before recreating the existing measurements. Thus, the effect of a change of a single input on the measurement results can be easily investigated.

The ability to recover input and recreate IMT measurements is useful for training an operator to use the image processing application 110. This capability enables experts to study measurement parameters specified by the operator and provide feedback. Operator-specified inputs may also be stored over a period of time and used to identify trends or changes in the specification of measured parameters by the operator, and ultimately to allow confirmation and verification of the operator's proficiency.

The adjustment module 166 may adjust the inputs and the results of the analysis of subsequent images in order to reduce computation time. This is particularly useful for tracking IMT values in a video clip of an ultrasound image comprising a series of images, where images preceding and images following any given image are similar to it. Given image similarity, the results of the necessary inputs and analysis do not typically change dramatically between successive images.

For example, the adjustment module 166 may adjust the user-selected regions 172, 178 into a continuous image. Other calculations discussed below may also be stored and reused by the adjustment module 166. For example, angle 100 and the location of adventitia, media, and lumen data provide a rough but still useful accurate estimate of the location of the boundary between tissue layers. The adjustment module 166 may also use reference values, or thresholds, generated from analysis of the histogram of the previous sampled region 178 for subsequent image measurements.

Referring to fig. 9, the adjustment module 166 may also adjust the inputs, data, and other results of the calculations to accommodate changes in the (acomod) image. For example, adjustment module 166 may move the location of region 172 according to the migration of the carotid artery between successive images due to movement of the artery itself or movement of sensor 64. That is, the module 166 may re-register the image to re-align objects in successive images of the same region.

Adjusting the position of the regions 172, 178 may provide a number of advantages, such as reducing the time spent manually selecting or automatically calculating the regions 172, 178. The reduction in computation time may improve the ability to track the location of tissue layers in the carotid artery in real time. By reducing the time spent in the calculation regions 172, 178, the image processing application 110 can measure video images at a higher frame rate without dropping or losing frames.

Accordingly, the adjustment module 166 may perform the adjustment process 186 automatically or with some degree of assistance or intervention by an operator. The analyzing step 188 is typically performed by other modules. However, this is the first step in the adjustment process 186. In identifying the lumen, adventitia, and lumen-in-image data, or both, analyzing 188 the first image may include calculating reference values for later calculation. Analysis 188 may also include locating the inter-tissue layer boundaries.

Once the first image has been analyzed, in analyzing the second image, applying 190 the analysis results to the second image may include using one or more of the same lumen, adventitia, or media data located in the first image. Applying 190 the results of the first image to the second image may include using the results only, without modification in the same manner as the results were applied by the first image. For example, the data calculated for the first image is applied unmodified to the second image. Alternatively, applying 190 may involve using the result as a coarse estimate (guess) that is then refined and modified during the measurement process.

For example, the adventitia 84 typically appears in the production image as the brightest portion of the carotid artery. Thus, locating the adventitia may involve finding the region of greatest intensity (brightness). Once the adventitia 84 is located in the first image, searching for the adventitia in the second image may be limited to a small region around the location of the adventitia in the first image. Therefore, by assuming that the adventitia 84 in the second image is in the vicinity of the same position as the adventitia 84 in the first image, the range in which the adventitia 84 is searched for in the second image is reduced. The positioning may be based on aligning the adventitia 84 in the two images, either automatically or with human assistance.

Applying 192 inputs to the second image may include using inputs manually or automatically provided for analyzing the first image for comparison with or directly for the second image. The analysis input to the first image may also be the result of the calculation by the adjustment module 166 for the adjustment step 194, described below. Thus, for example, the point 170 selected by the user for the first image may be used in the second image. Similarly, any height 174, 180 or width 176, 182 selected by the user, or automatically determined, for the measurement region 172 or the sampling region 178 may be used to analyze the new image, or to compare the second image.

Adjusting 194 the input to the second image may include determining how the second image differs from the first image. One method for making this determination may be to locate the adventitia 84 and indicate the location and/or orientation of identifiable irregularities in the first and second images. By comparing the position, orientation, or both of one or more points on the adventitia, which may be represented by the adventitia data, adjustment module 166 may calculate how the carotid artery is rotated or translated within the image. Such a point occurs where the carotid artery flares outward from a straight transition to the inflation point 90. Any translation and/or rotation may then be applied to the user-selected point 170 to specify the measurement region 172 and the sampling region 178. Rotation and/or translation may also be applied to any automatically determined position of the regions 172, 178.

Referring to FIG. 10, as well as FIGS. 6-9, the image labeling step 143 may calculate values that characterize a particular image for later use in image analysis. For example, the image identification module 114 may generate a histogram 200 of pixel intensities within the sampling region 178. The image identification module may also calculate adventitia, media, and lumen thresholds based on the histogram 200 to facilitate the location of image regions corresponding to the lumen 78, media 82, and adventitia 84.

For example, the lumen threshold 202 may be selected to be a suitable percentage (e.g., 10%) of the pixel intensity. Of course, other values may be selected depending on the characteristics of the image. Alternatively, the lumen threshold 202 may be selected based on the absolute range of pixel intensities within the sampling region 78. In some embodiments, the lumen threshold 202 may be calculated based on the minimum intensity 204 and the maximum intensity 206 that appear in the histogram 200. For example, the lumen threshold 202 may be calculated as a suitable fraction of the maximum difference in pixel intensity. The following formula proves to be effective:

lumen threshold (min strength + (fraction) × (max strength-min strength)

A fraction of 0.05 to 0.25 is effective and values of about 0.1 to about 0.2 are routinely successfully employed.

The adventitia threshold 208 may be hard coded to a fixed percentage (e.g., 90%) of pixels in intensity. The actual percentile chosen may be any suitable numerical value, depending on the image quality and the actual intensity of the pixels in the adventitia portion of the image. Alternatively, the adventitia threshold 208 may be equal to the maximum intensity 206. This choice is possible because the adventitia typically appears in the ultrasound image as the brightest band of pixels. In certain embodiments, the adventitia threshold 208 may also be selected at a fixed fraction of the maximum intensity difference below the highest intensity. The highest 5-25%, or other percentage, of the pixel intensity range may be used, with the highest 10% being conventionally used as a suitable threshold.

The media threshold 210 may be calculated based on the minimum intensity 204 and the maximum intensity 206. For example, the media threshold may be calculated according to the following formula:

the medium threshold is the minimum intensity +0.25 x (minimum intensity + maximum intensity), which is actually 25% of the total range of intensities.

Of course, other values for media threshold 210 are possible depending on the quality of the image and the actual intensity of pixels in the image portion of corresponding media 82. In some embodiments, the medium threshold 210 may be equal to the adventitia threshold 208. In some embodiments, the medium threshold 210 may be an intensity corresponding to a local minimum on the histogram 200 between the lumen threshold 202 and the adventitia threshold 208.

The image tagging module 114 may also receive and process input to enable a user to manually specify the thresholds 202, 208, 210. For example, the image tagging module 114 may enable a user to manually select a region of pixels within the lumen 78. The average or maximum intensity of the pixels in that region is then used as the lumen threshold 202. The user may similarly determine appropriate values for the adventitia threshold 208 and the media threshold 210 based on the maximum, minimum, or average intensities.

In some embodiments, image tagging module 114 may display histogram 200 and allow the user to select a threshold based on the communicated condition that the histogram pixel portion corresponds to the lumen, media, or adventitia. The image tagging module 114 may also display the histogram 200 and the ultrasound image of the carotid artery simultaneously, highlighting pixels below, or above the specific thresholds 202, 208, 210. The image tagging module 114 may then allow the operator to change the lumen threshold 202 and observe how the highlighted pixel region changes.

Referring to fig. 11, having determined the thresholds 202, 208, 210, the localization module 118 may analyze the lateral columns of pixels to locate the lumen 78, intima 80, media 82, adventitia 84, lumen/intima boundary, media/adventitia boundary, or any combination. The location module 118 may analyze lines of pixels oriented horizontally or at another angle based on the orientation of the carotid artery within the image. The positioning module 118 typically analyzes lines of pixels that extend substantially perpendicular to the tissue layer boundary.

Graph 218 is an example graph of pixel intensities versus their position within a column of pixels, where the horizontal axis 220 represents position and the vertical axis 222 represents pixel intensity. Starting from the left side of the graph 218, some important parts of the image 218 are: an inner cavity portion 224, which may be a portion below the inner cavity threshold 202; a lumen/intima boundary 226, which corresponds to the highest intensity gradient between the lumen portion 224 and the intima maximum 228; intimal maxima 228, corresponding to local maxima of the intima; a media portion 230 corresponding to a portion of the graph below media threshold 210; media dark pixels 231 generally provide local minima within media portion 230; a media/adventitia boundary 234, which is at or near the highest intensity gradient between the media dark pixel 231 and the adventitia maximum 236; and an adventitia maximum 236, which is generally the highest intensity pixel in measured region 172. It should be understood that the graphic 218 is representative of an idealized or typical image, but noise and weak contrast may cause the graphic of the pixel columns to appear different.

Referring to FIG. 12, the locating process 144 may include locating data to reduce the scope of searching for boundaries between layers of tissue. In one embodiment, the localization module identifies lumen data 240 and media data 242 that have a high probability of defining a lumen/intima boundary 244. The media data 242 and adventitia data 246 may be selected such that they have a high probability of defining a media/adventitia boundary 248. In certain embodiments, the localization process 144 may not localize the media data 242, but search for the lumen/intima boundary 244 and the media/adventitia boundary 248 between the lumen data 240 and the adventitia data 246.

The positioning process 144 may allow an operator to manually specify one, or all, of the data 240, 242, 246, or the boundaries 244, 248. Any method for manually specifying lines may be used to specify the boundaries 244, 248, or the data 240, 242, 246. For example, the operator may track the boundaries 244, 248 or the data 240, 242, 246 on the graphical display of the ultrasound image using an input device 22, such as a mouse 36. The user may establish the boundary 244, 248 or the data 240, 242, 246 by clicking on a series of points, which are then automatically connected to form a curve. Alternatively, the operator may establish the end points of the line and then establish the control points to define the curvature and the deflection points of the line (i.e., the Bezier curve). In other embodiments, the edges of the measurement region 172 may be used as the lumen data 240 or adventitia data 246. Referring to FIG. 13, the adventitia data locating process 146 may include locating 252 a starting adventitia pixel. The starting adventitia pixel may be found in a column of pixels centered on the user-selected point 170. Other suitable methods may include searching for the left or right most pixel column in the measurement region, or selecting the column at the center of the region selected by an automatic process. The adventitia 84 is typically the brightest portion of the image, so the absolute maximum intensity pixel may be searched to indicate the location of the adventitia 84.

Alternatively, the starting adventitia locating step 252 may include prompting the user to manually select a starting adventitia pixel. Yet another alternative is to search for the fewest neighboring pixels, each with an intensity above the adventitia threshold 208, and mark (e.g., label, identify, indicate) one of them as an adventitia pixel. This pixel will be used to fit the adventitia datum 246.

The adventitia locating process 146 may also include locating 254 adjacent adventitia pixels. Proceeding column by column, starting with the column adjacent to the starting adventitia pixel, the localization module 118 may search for the adjacent adventitia pixel in the remaining measurement region 172. The adjacent adventitia pixels may be located in a similar manner to the starting adventitia pixels. In some embodiments, adventitia pixels may be found in multiple sequences rather than moving from one column to an adjacent column. Sampling, periodic positioning, global maximum, left to right, right to left, etc. may provide starting points and may be affected by the sharpness and accuracy of the pattern.

With reference to FIG. 14, while continuing to refer to FIG. 13, the adventitia locating process 146 may compensate for noise and weak contrast by including a constraining step 256 and an extrapolating step 258. The constraining step 256 may limit the scope of searching for adventitia pixels to a small region centered or aligned (registered) relative to the lateral position of adventitia pixels in adjacent columns.

For example, adjacent columns of pixels may produce a series of patterns 260a-260 e. Graph 260a may represent the first column being analyzed. The maximum 262a is found and labeled as the adventitia pixel because it has the greatest intensity value. Constraining step 256 may limit the search for maximum 262b in graph 260b to a region 264 centered on or aligned with the location of maximum 262 a. The maximum falls outside this range and is then discarded because it is likely to be the result of noise. That is, the blood vessels are smooth. The adventitia does not wander extremely. The discarded maxima may then be excluded from the curve fit of any adventitia data 246.

The extrapolation step 258 may involve identifying a line 266 having a slope 268 that passes through at least two maxima 262a-262 e. Searching for other maxima may then be limited to a region 270, which is limited relative to the line 266. Thus, in the illustrated graph, the maximum 262c in graph 260c is not in region 270 and thus may be ignored. In some cases, the extrapolation step may involve ignoring a number of graphs 260a-260e whose maxima 262a-262e do not fall within the regions 264, 270.

As shown, graph 260d may have a maximum 262d outside of region 270, however, graph 260e has a maximum 262e within region 270. The number of columns that can be ignored in this manner can be adjusted by the user or automatically calculated based on image quality. In the case of poor image quality, the extrapolation step 258 may be made more quickly to find the appropriate (proper, clear) pixel column farther away. In some embodiments, the maxima 262a-262e used to establish the line 266 may be limited to pixel columns having good high contrast.

Referring again to FIG. 13, a curve fitting step 272 may establish adventitia data 246, a curve fit to the adventitia pixels located in these columns of pixels. Curve fitting is typically performed to smooth the adventitia 84 and compensate for noise that may not truly represent the adventitia 84. In one embodiment, the curve fitting step 272 may involve decomposing the measurement region into smaller segments (patches) and curve fitting each patch on a patch-by-patch basis. A function, such as a second order polynomial, a sinusoid or other trigonometric function, a power function, etc., may be selected to fit each segment. Each segment may be sized such that the path of adventitia pixels may be continuous, monotonic, have a single degree of curvature (e.g., no "S" shape in the segment), or have derivative continuity. Segment widths of 0.5 mm to 2 mm have been shown to provide a reasonable balance between accuracy, functional continuity, and computational speed.

Other embodiments are also possible. For example, wider segments may employ third-order polynomial interpolation to accommodate a greater likelihood of inflection points ("S" shapes) or derivative continuity in the adventitia pixel path. However, third order polynomial interpolation results in greater computational complexity and more computational time. Another alternative is to apply linear interpolation to very narrow segments. This can provide simple calculations, function continuity, but with discontinuities in the first derivative.

In certain embodiments, the fitted segment curves may overlap one another. This may provide the advantage that each curve-fitted segment has a substantially matching slope with the adjacent segment at the point of adjacency. However, this approach may introduce complex calculations by requiring many pixels to be analyzed twice. However, it provides a relatively simple calculation for each segment.

Referring to fig. 15, the lumen data localization process 148 may identify the portion of the image corresponding to the lumen 78 and establish the lumen data 240. The lumen data locating process 148 may include locating 280 a low intensity region. This step may include finding a specified number of adjacent pixels in a column of pixels below the lumen threshold 202. The search for a segment (band) of low intensity pixels typically begins at the adventitia 84 and proceeds toward the center of the lumen 78. A four pixel wide area has proven to be suitable. Alternatively, step 280 may include searching for a neighboring group of pixels whose average intensity is below the lumen threshold 202.

After the low intensity region is located, the next step may be a confirmation step 282. In some cases, dark regions within the intima/media region may be large enough to have four pixels below the lumen threshold 202. Thus, the low intensity region may be identified to ensure that it is indeed within the lumen 78. One method of confirmation is to ensure that the low intensity region is adjacent to the large intensity gradient, which is typically at the lumen/intima boundary 226. The proximity of the intensity gradient required to identify the low intensity regions may vary.

For example, the confirmation option may require that the low intensity region be directly adjacent to the large intensity gradient. Alternatively, it is recognized that only the low intensity region may be required to be within a specified number of pixels (e.g., distance) from the high intensity gradient. When the low intensity region has a high probability of being invalid, the lumen data locating process 148 may repeat, starting at the location of the invalid low intensity region during the first iteration, and moving away from the adventitia 84.

The lumen data positioning process 148 may also include a compensation step 284. In some cases, it is difficult to verify whether the low intensity region is adjacent to the lumen/intima boundary 224 because the ultrasound imaging process has limitations that tend to fail to capture intensity gradients, leaving behind a weakly contrasting image region. Accordingly, the compensation step 284 may include methods to compensate for contrast deficiencies for boundaries in the weak contrast regions by extrapolation, interpolation, or both. The confirmation 282 may thus include verifying the proximity of the low intensity region to the extrapolated or extrapolated boundary.

The curve fitting step 286 may include the found low intensity regions into the lumen data 240. In certain embodiments, the path that includes the first pixel found in each column of low intensity regions is a curve that is fitted to establish the lumen data 240. In embodiments where intensity averages are used to locate the lumen, the centered (dimensionally) located pixel in the set of averaged pixels may be used to curve fit the lumen data 240. The curve fitting step 286 may curve fit the pixel path in conjunction with the adventitia data locating process 146 in the manner discussed above.

FIG. 16 illustrates a method for performing the optional low contrast compensation step 284, which includes an identification step 288, a constraint step 290, a bridging step 292, and a verification step 294. The identifying step 288 may identify portions of the measurement region 172 that appear to be of high quality. Identifying 288 may include identifying a horizontal region that is at least three to five pixel columns wide, with each column having a relatively high contrast. The larger or smaller horizontal regions may be selected based on the type and quality of the image, and in some embodiments, the contrast may be determined by searching for the largest, or sufficiently large, intensity gradient in a column. The gradient values sufficient to identify a column of pixels as "high contrast" may be fixed, user selected, automatically selected, or a combination thereof, or all of the above, based on the characteristics of the image. In some embodiments, the required intensity gradient may be a certain percentage of the maximum intensity gradient found in the sampling region 178.

Identifying 288 can also include verifying that the large intensity gradient of each column in the horizontal region occurs at about the same location in the columns, deviating from each other by no more than a predetermined number of pixels. Thus, for example, if a high intensity gradient is located at the 7 th pixel in a column of pixels, identification 288 may include verifying that the high intensity gradient occurs somewhere between the 70 th and 80 th pixels in adjacent columns. Columns whose intensity gradients do not fall within this region may be excluded from the horizontal region of high intensity pixels in order to assess the quality of the image, and extrapolate and interpolate the gradient or boundary locations. In some embodiments, only regions of a particular width having a high intensity gradient adjacent a column, and the high intensity gradient occurring at approximately the same lateral position, are considered high contrast regions.

The constraining step 290 may attempt to identify the location of the lumen/intima boundary 226, or more generally any boundary or feature, in the absence of high contrast. One way to achieve this is to limit the search area. Constraint 290 may therefore search for the maximum gradient in a low contrast region between two high contrast regions by limiting the search to regions centered on a line drawn from a large intensity gradient over one high contrast region to a second large intensity gradient.

Constraints 290 may also include defining boundaries using different values. However, in high contrast regions, a significant value may be used to identify, limit, or specify which gradient represents a boundary. Constraint 290 may include determining the maximum intensity gradient in a fairly low contrast region and using some percentage of this smaller value to define which gradient is large enough to represent a boundary. Similarly, constraining 290 may include finding the steepest gradient above some minimum in the constrained region.

The bridging step 292 may be based on the location of the boundary or the gradient in the much higher contrast region on one side, including interpolating the location of the boundary or other intensity gradients in the low contrast region. Alternatively, in some embodiments, the position or gradient of the boundary may be extrapolated to one side of the low contrast region based on the position of the high contrast region.

The verification step 294 may verify that the high contrast regions are suitable for proving that the push-in and push-out to much lower contrast regions are correct. The verification step 294 can include comparing the number of columns in these "high" contrast regions with the number of columns in the "low" contrast regions. Extrapolation and interpolation cannot improve accuracy when there are more pixel columns in the low-contrast area than in the high-contrast area.

Referring to FIG. 17, in an alternate embodiment, the lumen data positioning process 148 includes a translation step 300 and a translation validation step 302. Referring to FIG. 18, the translating step 300 may include translating the adventitia datum 246 toward the center of the lumen 78. The translation verification step 302 may average the intensities of all pixels located on the path of the translation. The average, median, or some number of total pixels corresponding to an intensity less than the lumen threshold 202, or some other minimum, the translation verification step 302 may include establishing the translated adventitia datum 246 as the lumen datum 240. Alternatively, the translation verification step 302 may include labeling the translated adventitia datum 246 as the lumen datum 240 only where all pixel intensities on the translated datum 246 are below the lumen threshold 202 or other minimum. Alternatively, the translated adventitia data 246 may be used only as a starting point for another curve fitting process, or as a center, edge, or other identifying point, for a region of the constrained search lumen data 240, for example, the method of FIG. 15 may be used to search the lumen data 240 within the constrained search region.

Referring to FIG. 19, the media data positioning process 150 may include a positioning step 308 and a curve fitting step 310. The locating step 308 may identify a media dark pixel path that is subsequently suitable for generating the media data 242. In accordance with the present invention, curve fitting step 310 may curve fit the media dark pixel path to generate media data 242 in the same manner as the other curve fitting steps already discussed. In some embodiments, the media data location process 150 may be eliminated and the lumen data 240 may be used wherever the media data 242 is used to limit the scope of the search. In another embodiment, the lumen data 240 may be eliminated and only the adventitia 84 may constrain the search for boundaries between tissue layers. For example, the search medium/adventitia boundary may only begin with adventitia data 246 and move toward lumen 78.

Referring to FIG. 20, a locating process 312 illustrates one method for locating 308 a media dark pixel path. The process 312 may be performed on columns of pixels in the measurement region 172. The process 312 may begin by examining pixels located on, or near, the adventitia datum 246. After determining 314 the pixel intensity, the process 312 may determine 316 whether the pixel is a local minimum. If so, the process 312 may determine 318 whether the pixel density is less than the media threshold 210. If so, the pixel is marked 320 or indicated 320 as a media dark pixel, and process 312 is performed on another column of pixels.

If the pixel is not a local minimum, the process 312 may determine 322 if the pixel is located less than a predetermined distance from the adventitia 84. Suitable values for this distance may be about one-half to two-thirds of the distance from the outer membrane 84 to the inner cavity 78. The adventitia data 246 and the lumen data 240 may be used to specify the location of the adventitia 84 and the lumen 78 in order to determine the distance therebetween. If the pixel is less than the specified distance from the adventitia 84, then the process 312 moves 323 to the next pixel in the column, in some embodiments, away from the adventitia 84, and the process 312 is repeated. If the pixel is spaced a prescribed distance from the lumen 78, it is labeled 324 as a media dark pixel and process 312 is performed for any remaining columns of pixels.

If the minimum value of the intensity is not less than the media threshold 210, the process 312 may determine 326 whether the distance from the pixel to the adventitia 84 is greater than or equal to the same predetermined value as in step 322. If greater than or equal to this value, the corresponding pixel is marked 328 as a media dark pixel and process 312 is performed on any remaining columns. If less than this value, the process 312 branches 329 to another pixel point in the column, typically to the adventitia 84, and the process 312 is repeated.

FIG. 21 illustrates another embodiment of a media data location process 150. The minimum location step 330 may look for a local minimum of intensity in each column of pixels between the adventitia 84 and the lumen 78. In certain embodiments, the minimum location step 330 may search for a local minimum between the adventitia data 246 and the lumen data 340. The minimum location step 330 may begin with the adventitia datum 246 and shift the search minimum to the lumen datum 240. In columns of pixels with weak contrast, the minimum location step 330 may include extrapolating or interpolating the possible locations of local minima representing the medium to one side of the column or columns of pixels with weak contrast based on the location of the verified minimum on either side. The possible locations of local minima determined by extrapolation or interpolation may be used as the locations of dark pixels in the column instead of the actual valid local minima.

Once the minimum is found, the validation step 332 may verify that the minimum is likely located within the medium 82. The minimum value may be confirmed 332 by ensuring that it is below the media threshold 210. The minimum may also be confirmed 332 by ensuring that it is adjacent to a high intensity gradient between the minimum and the adventitia datum 246, since the darker medium 82 is adjacent to the much lighter adventitia 84, and thus will have an intensity gradient between them. Validation step 332 may include marking valid minima as media dark pixels for use in creating media data 242. If an inappropriate minimum is found, the process 150 may repeat, beginning with the inappropriate minimum, shifting toward the lumen 78.

The confirming step 332 may also include detecting the location of the minimum. The validation 332 may ensure that only those minima within a specified distance from the adventitia 84 are marked as media dark pixels, which are used to calculate the media data 242. The operable value for the prescribed distance may be between about one-half to about two-thirds of the distance from the outer membrane 84 to the inner cavity 78. If none of the minima fall below the media threshold 210, lie near a large intensity gradient, or both, then confirmation 332 may include marking the pixels a specified distance from the adventitia as media dark pixels, which are used to calculate (curve fit) media data 242.

Referring to FIG. 22, and also to FIG. 21, the curve fitting step 334 may establish temporary media data 336 that includes a curve fit of media dark pixels 338 located laterally within each column of pixels on the image domain (domain). The curve fitting step 334 may use any of the curve fitting methods discussed above in conjunction with other data or other suitable methods.

Adjustment step 340 may change the location of media dark pixels 338 used to calculate media data 242. For example, each media dark pixel 338 can be checked to see if it is located between temporary media data 338 and adventitia data 246. The medium 82 has no actual intrusion into the outer membrane 84. Media dark pixels between the outer film data 246 and the temporary media data 338 may be moved to the temporary media data 336 or replaced by dots or pixels at the temporary media data 336. A curve fitting step 342 may then curve fit the media data 242 to the modified set of media dark pixels 338. The curve fitting step 342 may use any of the curve fitting methods discussed above in conjunction with other data or other suitable methods. Alternatively, the temporary media data 338 may be used as the media data 242 itself.

Fig. 23 illustrates a lumen/intimal border definition process 152. The define step 346 defines 346 the searched range. For example, in one embodiment, the search range is limited to the region between the lumen data 240 and the media data 242. Defining 346 the search range may include searching only the region between the lumen data 240 and the first local maximum found when searching from the lumen data 240 to the media data 242. Some embodiments may require that the local maxima have an intensity above the media threshold 210. In certain embodiments, defining 346 the search range may include manually or automatically adjusting the location of the lumen data 240 and/or the media data 242. For example, the user may click on the graphical representation of the lumen data 240 and translate it laterally to different positions to observe the quality or correspondence of the fit. In other embodiments, the search range may be defined 346 as the region between the adventitia datum 246 and the edge of the measurement region 172 located within the lumen.

In certain embodiments, the operator may select a point on the proximal lumen/intima boundary 244. Defining 346 the scope of the search may include searching only a small area centered on the operator selected point, or a line interpolated between the operator selected points. Alternatively, the operator or software may select or specify a point, or series of points, that is well within the lumen 78 to define one boundary of the search area.

The locating step 348 may begin with the lumen data 240, or other bounding boundary, such as the edge of the measurement region 172, and search the media data 242 for a maximum positive intensity gradient. In embodiments where the media data 242 is not located, the locating step 348 may search for the largest positive intensity gradient from the lumen data 240, or other boundary, to the adventitia data 246. The validation step 350 may verify that the gradient may represent the lumen/intima boundary 244. In certain embodiments, the locating step 348 may involve searching for the largest negative intensity gradient when transitioning from the adventitia data 246, or a local maximum above the media threshold 210, to the lumen data 240, or media data 242. In certain embodiments, the validation 350 may include discarding gradients where the gradient-defining pixels are below a prescribed threshold, such as the lumen threshold 202.

The defining step 346, the locating step 348, and the confirming step 350 may be repeated until a maximum (steepest) effective intensity gradient is found. Defining 346 the search range may therefore include limiting the search range to those columns for which no pixel has been examined so far. For example, defining 346 the search range may include limiting the search range to a region between the invalid gradient and the media data 242 or the first local maximum above the media threshold 210.

An optional specification step 352 may enable an operator to manually specify the location of the lumen/intima boundary 244 at one or more points. As discussed above, the optional compensation step 284 may extrapolate or interpolate the location of the lumen/intima boundary 244 within the relatively low contrast region based on the portion of the lumen/intima boundary 244 found within the relatively high contrast region. The compensation step 284 may also extrapolate or interpolate between operator specified points and high contrast areas.

FIG. 24 illustrates one embodiment of a media/adventitia boundary locating process 154. The qualifying step 358 may qualify the scope of the search. For example, in one embodiment, the search range is limited to the pixel columns of the portion between the media data 242 and the adventitia data 246. Alternatively, the search range is limited to the region between the lumen data 240 and the adventitia data 246. In another embodiment, the search range may be defined 358 as the region between the adventitia datum 246 and the edge of the measurement region located within the lumen. In certain embodiments, defining 358 the search region may also include manually or automatically translating the media data 242, the adventitia data 246, or both. In another embodiment, the search range may be limited to a region between media data 242 and a local maximum having a corresponding intensity above adventitia threshold 208 or other minimum.

The locating step 360 may identify the largest positive gradient within the search range. The positioning step 360 may involve examining each pixel, starting from the media data 242, or other boundaries, such as the edges of the measurement region 172 or lumen data 240, and transferring to the adventitia data 246. The validation step 362 may verify that the intensity gradient is likely to be the media/adventitia boundary 248. When the gradient-defining pixels are below a certain value, such as media threshold 210, the confirmation 362 may include discarding the gradient.

When the gradient is discarded in the validation step 362, the qualifying step 358, the locating step 360, and the validation step 362 may be repeated to find and validate the next maximum intensity gradient until the maximum valid intensity gradient is found. The defining step 358 may also include limiting the search range to the region between the media data 242, or other boundary, and the location of the invalid intensity gradient. Alternatively, the specification step 364 may enable an operator to manually specify the approximate location of the media/adventitia boundary 248 at one or more points. As described above, the compensation step 284 may extrapolate or interpolate the position of the media/adventitia boundary 248 in the contrast region based on the portion of the media/adventitia boundary 248 found along the media/adventitia boundary 248 at the high contrast region and/or at the operator-specified point.

Referring to fig. 25, the calculation module 120 may calculate an IMT value based on the distance between the lumen/intima boundary 244 and the media/adventitia boundary 248. In certain embodiments, the calculation module 120 may calculate the distance between the lumen/intima boundary 244 and the media/adventitia boundary 248 for each column of pixels and average them together to produce a final value. The calculation module 120 may also convert the calculated IMT value to its actual, real-world value based on the calibration factor calculated by the calibration module 112.

In certain embodiments, the calculation module 120 may remove (reject) peaks or other discontinuities in slope at the media/adventitia boundary 258. For example, the calculation module 120 may also search for peaks whose height is a specified multiple of their width. For example, a peak having a height that is three times (or other effective multiple) the width of the base may be identified. The portion of the media/adventitia boundary 248 that forms the peak may be replaced with an average of the boundary positions on either side of the peak. The calculation module 120 may also remove the peak from the lumen/intima boundary 244.

The calculation module 120 can also curve fit one or both of the media/adventitia boundary 248, the lumen/intima boundary 244. In certain embodiments, the calculation module 120 curve fits the boundaries 244, 248 after removing the peaks from the boundaries 244, 248 so that the substantially erroneous data does not affect the final curve fit.

The calculation module 120 may include a slope compensation module 370. The slope compensation module 370 may adjust the IMT measurement for the angle 100 of the carotid artery relative to the horizontal direction 74. For example, in some embodiments, the slope compensation module 370 may multiply the IMT measurement by the cosine of the angle 100. The angle 100 may be calculated by fitting a line to the lumen/intima boundary 244, the media/adventitia boundary 248, or a line of pixels at a midpoint between the lumen/intima boundary 244 and the media/adventitia boundary 248 for each column of pixels. The angle 100 may be set equal to the angle of the line relative to the horizontal direction 74. Alternatively, the angle 100 may be calculated with a line fit to one or a combination of the lumen data 240, the media data 242, and/or the adventitia data 248. In certain embodiments, the angle 100 may be calculated based on a line connecting the leftmost and rightmost points, including the lumen data 240, the media data 242, the adventitia data 240, the lumen/intima boundary 244, or the media adventitia boundary 248. Alternatively, the operator may select two points, which the slope compensation module 370 may then use to define the angle 100.

The calculation module 120 may also include a taper compensation module 372 for adjusting the IMT measurements to offset any taper in IMT thickness that may have an effect on the measurements. One method for eliminating this type of variation is to measure the IMT without a cone-shaped area of influence. For example, the IMT of the segment 98 located at a distance between 10 mm and 20 mm from the flared portion 90 is typically not significantly tapered.

The taper compensation module 372 may locate the bifurcation by searching for the expansion point 90. In one embodiment, the taper compensation module 372 fits a straight line to a substantially straight portion of the outer membrane 84. The taper compensation module 386 may then extrapolate the tine (tine) to the branch, checking the intensity of the pixels located on the line. The line extends into the lumen 78 when the pixels falling on the line have always had an intensity below the lumen threshold 202. The location where the line encounters a low intensity pixel will correspond to a location near the dilation point 90 and near the branch. Of course, different methods may be used to locate the expansion point 90.

Referring to FIG. 26, while also referring to FIG. 25, taper compensation module 372 may perform a taper compensation process 374. The taper compensation process 374 may include generating 376 an IMT database 133. Referring to FIG. 27, generating 390 the IMT database 133 may include measuring IMT of the carotid artery at different subsections (subsections)378 and recording an average IMT for each subsection along each subsection location. In certain embodiments, the IMTs of the different subsections 378 may be curve-fitted and are polynomials or other mathematical expressions that are a curve fit of the records. Sub-segment 378 generally spans sections 94, 98, or portions of both segments 94 and 98, so as to include a tapering effect near expansion point 90. The width of the sub-segments 378 may correlate the degree of taper to regions of large angular taper divided into narrower sub-segments 378. The IMT database 133 typically includes measurements for a large number of patients.

Studies have shown that the degree of tapering depends primarily on the average IMT, with arteries having a smaller average IMT having a lesser degree of tapering than arteries having a larger average IMT. Accordingly, based on the IMT at the point of normalized distance from the inflation point 90, generating 376 an IMT database may include indexing each series of measurements acquired from the ultrasound image. For example, because the segment 98 extending 10 mm to 20 mm from the expansion point 90 has a substantially constant IMT, the measurements taken from the image may be indexed by the IMT of the point 15 mm from the expansion point 90. Alternatively, a regional average IMT centered at or near a 15 mm point may be used.

Furthermore, IMT measurements for multiple patients with similar normalized point IMTs may be averaged together and the averaged results stored for later use and indexed by their average IMTs at the normalized points. Typically, the IMT measurements for sub-segment 378 at a specified distance from the inflation point 90 for one patient are averaged with the IMT measurements for sub-segment 378 at the same distance in the ultrasound image of another patient.

Referring to fig. 28, and also to fig. 26 and 27, calculating 380 the normalization factor may include retrieving from the IMT database 133 IMT measurements 136 obtained from a carotid artery, or carotid arteries, that are ultrasound images of the current same point, the IMT measurements 136 having substantially the same IMT thickness. Thus, for example, if the current ultrasound image has an IMT of 0.27 millimeters at a distance of 15 millimeters from the inflation point 90, then calculating 380 the normalization factor may include restoring the IMT measurement 136 for the artery having an IMT of 0.27 millimeters at the corresponding point. Alternatively, IMT measurements 136 may be recovered for the recorded arterial measurements, having an IMT value at a normalized point defining the IMT of the current artery at that point.

Based on the stored IMT measurement 136 or a subsection 378 of the measurement 136, a normalization factor may be calculated 380. For example, sub-segment 378a may have an IMT 382 and be located at point 384. Sub-segment 378b may have an IMT 386 and be located at another point 388. The distance of the point 388 from the expansion point 90, which is used to normalize nearly all IMT measurements 136, may be selected as a normalized distance. By dividing IMT 386 by IMT 382, a normalization factor may be calculated for sub-segment 378 a. In a similar manner, for each sub-segment 378, IMT 386 may be divided by IMT to calculate 380 a normalization factor for each sub-segment 378.

Referring again to FIG. 26, applying 390 the normalization factor may include multiplying the normalization factor by the IMT of the subsection 378 in the current ultrasound image corresponding to the normalization factor. Thus, for example, the subsection 378 centered 7 millimeters from the expansion point 90 in the current image would be multiplied by a normalization factor calculated at a distance of 7 millimeters from the expansion point 90. In this manner, as shown in FIG. 29, the IMT at each subsection 378 in graph 392 is transformed to be approximately equivalent to the IMT at the normalized point 388 in graph 394. The normalized IMT of each subsection 378 may then be averaged to produce a final value that may be reported.

A number of alternative ways of applying the normalization factor are possible. For example, rather than dividing the current ultrasound image into sub-segments 378, a normalization factor may be applied to the IMT of each column of pixels. The interpolation between normalization factors calculated for sub-segment 378 centered at a position defining a column of pixel horizontal positions may be used to normalize IMT for a single column of pixels. Alternatively, in the recovered IMT measurements 136, a normalization factor may be calculated for each column of pixels. In another embodiment, a mathematical expression of the stored IMT measurements 136 is used to calculate a normalization factor at each column pixel location.

Referring again to FIG. 25, the calculation module 120 may also include a data scaling module 398 and a diagnostic module 400. The data conversion module 398 may compile and statistically analyze IMT measurements and other data to derive the diagnostic data 131. The diagnostic module 400 may retrieve the diagnostic data 131 in order to correlate the patient's IMT with the patient's cardiovascular disease risk.

As can be appreciated from the foregoing, imaging devices, such as ultrasound imaging devices, computer-aided topography imaging devices, magnetic resonance imaging devices, and the like, can accurately capture fine details of human anatomical features. By using the imaging device in conjunction with measurement techniques, tissue density in terms of tissue structure, such as "plaque" associated with an arterial wall, can be determined (plaque is defined as an abnormally thickened local region of any portion of the arterial wall).

For example, it is necessary to determine the tissue density in the carotid artery wall. The presence of carotid plaque has been shown to be a good indicator of the presence and extent of atherosclerosis in the entire vasculature.

The density of the textural aspects may vary depending on the characteristics of the textural aspects. The tissue structure has special features, such as common sense that can be determined from the density associated therewith is useful in condition diagnosis and/or can provide some information. For example, the plaque may be dense and significantly calcified, or may be floppy and non-calcified, or may be calcified to any degree. Softer, non-calcified plaques are generally the more dangerous type of plaque as they are more prone to rupture/dislodgement and cause an occlusion or clot. It is therefore important to be able to determine the presence of these softer plaques and to determine the density of these plaques. Although there are techniques for non-invasively determining the presence of calcified plaque, such as CAT scanning, in addition to ultrasound, these techniques often have difficulty in finding non-calcified plaque. The ultrasound technique can detect dense calcified plaque and softer non-calcified plaque. Embodiments of the present invention utilize images, such as those available from ultrasound devices or other suitable imaging devices, to locate and characterize the density of carotid arteries and/or branch vessel plaques associated therewith.

Digital images, such as those provided by ultrasound devices, are typically composed of rows and columns of digitally sampled values representing the tissue density at each location in the image. These sampled values are referred to as pixel values, or simply pixels. Brighter pixel intensities correspond to denser tissue. The measurement process for measuring plaque density in the carotid artery according to embodiments of the present invention uses these pixel intensities, and their relative positions in the digital image, to determine plaque size, location, and/or density.

FIG. 30 illustrates a system for providing density information in terms of tissue structure according to an embodiment of the present invention. In the illustrated embodiment, ultrasound imaging device 3001 captures images including tissue structures of interest, such as carotid artery images, and provides the images to plaque density measurement mechanism 3003 via communication medium 3002. For example, the plaque density measurement mechanism 3003 can comprise a general purpose computer (e.g., a personal computer) system operable under the control of an instruction set to define the operations described herein. Alternatively, the plaque density measurement mechanism 3003 may be integrated into imaging device 3001, such as may comprise a processor operable under the control of a set of instructions to define the operations defined herein and/or one or more Application Specific Integrated Circuits (ASICs). Communication medium 3002 may include digital or analog interfaces, such as may include network links (e.g., a Wide Area Network (WAN), a Metropolitan Area Network (MAN), a Local Area Network (LAN), a Public Switched Telephone Network (PSTN), the internet, an intranet, an extranet, a cable transmission system, and/or the like), parallel data links (e.g., Centronics parallel interfaces, IEEE 1284 parallel interfaces, IEEE 488 parallel interfaces, Small Computer Systems Interconnect (SCSI) parallel interfaces, Peripheral Component Interconnect (PCI) parallel buses, and/or the like), serial data links (e.g., RS232 serial interfaces, RS422 serial interfaces, Universal Serial Bus (USB) interfaces, IEEE1394 serial interfaces, and/or the like).

The plaque density measurement mechanism 3003 identifies tissue structures in the ultrasound image, such as plaque regions in the carotid artery, and determines their density. For example, the plaque density measurement mechanism 3003 may create data to identify regions of plaque as described above. Additionally or alternatively, the user input may be used to identify regions of the plaque, plaque boundaries, plaque centers, and the like. After the plaque regions are identified, the plaque density measurement mechanism 3003 of the preferred embodiment determines plaque density by analyzing the image intensity within the plaque regions. For example, the plaque density measurement mechanism 3003 can normalize the intensity of the entire image, or certain portions thereof, and then use the average or median pixel intensity values associated with the plaque area. Additionally or alternatively, the plaque density measurement mechanism 3003 can analyze different portions of the plaque region, such as different tissue layers, regions adjacent to other tissue structures, and so forth, to determine a density associated therewith. Thus, several plaque densities can be determined relative to the identified plaque regions. The plaque density measurement mechanism 3003 can additionally or alternatively determine associated information such as the size of the plaque region, the thickness of the plaque region, the relative size (e.g., percentage) of the plaque region to another aspect of the tissue structure or the tissue structure itself, the relative position of the plaque region to another aspect of the tissue structure or the tissue structure itself, and/or the like.

The plaque density measurement mechanism 3003 of the illustrated embodiment provides the aforementioned plaque density information to the plaque density report generator 3004. Plaque density report generator 3004 can comprise a general purpose computer (e.g., a personal computer) system operatively controlled by an instruction set defining operations described herein. Alternatively, plaque density report generator 3004 may be integrated into imaging device 3001, such as may comprise a processor and/or one or more ASICs, the processor being operatively controlled by an instruction set defining operations defined herein. The plaque density report generator 3004 of the preferred embodiment provides some form of output of plaque density information that is useful in diagnosing a condition of a tissue structure, or is otherwise useful. For example, the plaque density report generator 3004 may compare plaque density values, thicknesses, sizes, locations, etc. to a database of plaque density values, thicknesses, sizes, locations, etc. to report the relative condition of the patient. Additionally or alternatively, the plaque density report generator 3004 may compare plaque density values, thicknesses, sizes, locations, etc. to historical information associated with a particular individual to determine trends, such as the growth rate of plaque, the calcification rate of plaque, etc.

Any or all of the foregoing information generated by the plaque density report generator 3004 may be reported to a user or operator in any of a variety of formats, using a variety of mediums. For example, plaque density report generator 3004 may display information in real time or near real time on plaque density display device 3005, which display device 3005 may include, for example, a display monitor (e.g., a Cathode Ray Tube (CRT), Liquid Crystal Display (LCD), plasma display, etc.), which may be part of or separate from ultrasound imaging device 3001. Atheroma density report generator 3004 may output information in the form of a hardcopy on atheroma density printing device 3007, which may include, for example, a printer (e.g., dot matrix printer, laser printer, page printer, ink jet printer, etc.) coupled to atheroma density report generator 3004 via communication medium 3006. Communication medium 3006 may include any of a variety of communication media such as those described above with respect to communication medium 3002. Additionally or alternatively, plaque density report generator 3004 reports the foregoing information to a user or operator, which information may be stored in one or more databases and/or transmitted to one or more systems by embodiments of the present invention.

Having described an embodiment of the organizational structure aspect density information system, reference is now made to FIG. 31, which illustrates a flowchart of the operation of the organizational structure aspect density information system, according to an embodiment. At step 3101 of the illustrated embodiment, the patient provides the tissue structure to be analyzed, which is shown as the carotid artery in the illustrated embodiment. At step 3102, an ultrasound imaging device is used to capture a digital image of the carotid artery. Then, at step 3103, a digital image of the carotid artery is provided to the plaque density measuring device of the present invention.

In step 3103, the plaque density measurement device operates to locate the location of carotid arteries or other anatomical aspects of interest within the ultrasound image (plaque in the illustrated embodiment). There are several ways in which this can be achieved. One mechanism is for the operator to track the boundaries of the plaque. This is the simplest implementation mechanism, as it does not require automatic plaque edge determination. Another way to locate atherosclerotic plaques is for the operator to specify one or more points within a region of interest in the carotid artery wall. The algorithm may then search outward from these points until the plaque boundary is determined. For example, one embodiment determines an atherosclerotic plaque boundary from the identified points by finding the maximum intensity gradient in all directions from the specified points. Another way to locate the plaque is to have the operator prescribe an area or path around the exterior of the plaque in a way that ensures that each pixel making up the plaque is completely surrounded by the prescribed area. The algorithm then searches for the plaque boundary by searching inward from the boundary, looking for the maximum intensity gradient.

Once the plaque is located and its outer boundary is determined, several density features may be calculated by the plaque density measurement device at step 3103. These features include, but are not limited to, mean density, peak density, density range (maximum density minus minimum density), total plaque area, area of densest region, density histogram, density standard deviation or sum of variance, but are not limited to density centroid.

It should be understood that the intensity values that make up an ultrasound image are relative intensity values. Thus, the actual pixel intensity values may be adjusted to be more or less sensitive to changes in tissue density. Depending on the image gain setting selected by the operator, the image intensity may change sharply even for the same image. In this way, it is necessary to normalize the pixel intensities as much as possible for these operator influences by normalizing the collected density information for a given plaque. This can be achieved by modifying each density value collected by an adjustment factor that depends on the known tissue density of the landmark, with a relatively constant density for different patient landmarks. One landmark of known density is the lumen. Measuring the density of the lumen measures the density of the blood flowing in the artery. The blood density is assumed to be constant for different patients. The density of the lumen gives a lower bound to the tissue density found in the image. Another landmark used as an upper limit of tissue density is the adventitia. The adventitia tends to be the most dense tissue in the carotid artery image. The density is also relatively constant from patient to patient. By normalizing the plaque density on a scale from luminal to adventitial density, a gain factor that is controlled by the operator and directly affects the tissue density metric can be normalized.

One mechanism for normalizing these two known tissue densities is simple linear scaling and compensation. For example, assuming that the lumen average density is 10 and the adventitia average density is 170, for normalized plaque density, the value is 77, and thus the percentage density over the range from the lumen density to the adventitia density can be determined. This can be calculated by the following equation:

normalized tissue density (77-10)/(170-10) 0.419-41.9%

Specifying normalized tissue density as a percentage of lumen density to adventitia density is only one option. There are many other standardized reporting methods. For example, embodiments may utilize a percentage from the inner membrane/media density to the outer membrane density.

By using a normalization procedure, such as one or more of the procedures described above, the plaque density is specified over a range that is independent of the instrument gain setting. At step 3104, the plaque density data determined by the plaque density measuring device is provided to a plaque density reporting process. The plaque density reporting process outputs the plaque density information in any of a variety of formats, such as by displaying the plaque density report at step 3106 and/or printing the plaque density report at step 3107.

The foregoing plaque density report may provide plaque density information in a format that is readily understood and interpreted by a user. For example, one embodiment of the present invention outputs a number from 1 to 5, which represents the severity of a given atherosclerotic plaque stiffness. These five numbers (bins) will be associated with five ranges of densities that correspond to risk levels, such that the risk levels can be determined by collecting plaque density values from a large population that is combined with a history of events (heart attacks, stroke, etc.).

In another embodiment of plaque density usage, generating a database of normalized plaque values allows risk stratification based on plaque density to characterize plaque. By having the database contain a large number of plaque densities from a large and diverse population combined with the occurrence of an event, the database can then be used in conjunction with individual plaque density measurements for screening purposes.

Although the present invention and its advantages have been described in detail, it should be understood that various changes, substitutions and alterations can be made herein without departing from the invention as defined by the appended claims. Moreover, the scope of the present application is not intended to be limited to the particular embodiments of the process, machine, manufacture, composition of matter, means, methods and steps described in the specification. As one will appreciate from the disclosure, processes, machines, manufacture, compositions of matter, means, methods, or steps, presently existing or later to be developed that perform substantially the same function or achieve substantially the same result as the corresponding embodiments described herein may be utilized. Accordingly, the appended claims are intended to include within their scope such processes, machines, manufacture, compositions of matter, means, methods, or steps.

Claims (49)

1. A method for characterizing a given anatomical aspect in an image, the method comprising:

analyzing the image to identify the aspect of the tissue structure, wherein analyzing the image comprises: analyzing the columns of pixels to locate at least one of a lumen, intima, media, adventitia, a lumen/intima boundary, and a media/adventitia boundary;

normalizing the intensity information by applying an adjustment factor to intensity information of the image, the adjustment factor depending on a known density of one or more landmarks in the image;

determining density information for the aspect using the normalized strength information; and

characterizing said aspects with said density information, wherein a database comprising a plurality of plaque densities from a plurality and different populations associated with the occurrence of an event is used in conjunction with the individual plaque density measurements for screening.

2. The method of claim 1, wherein said analyzing said image comprises:

analyzing intensity gradients within the image.

3. The method of claim 1, wherein said analyzing said image comprises:

the data is identified with image intensity information.

4. The method of claim 1, wherein said analyzing said image comprises:

tissue boundaries are identified using image intensity information.

5. The method of claim 1, wherein the one or more landmarks have a relatively constant density for different images.

6. The method of claim 5, wherein the one or more landmarks include a blood vessel lumen.

7. The method of claim 5, wherein the one or more landmarks include an adventitia.

8. The method of claim 1, wherein the determining density information comprises:

determining an average intensity of the image within the region identified as the aspect so as to provide an average density.

9. The method of claim 1, wherein the determining density information comprises:

determining a peak intensity of the image within the region identified as the aspect to provide a peak density.

10. The method of claim 1, wherein the determining density information comprises:

the maximum intensity within the region identified as the aspect minus the minimum intensity within the region identified as the aspect is determined to provide a density range.

11. The method of claim 1, wherein the determining density information comprises:

an intensity histogram is determined for intensities within the region identified as the aspect to provide a density histogram.

12. The method of claim 1, wherein the determining density information comprises:

the standard deviation of the intensity within the region identified as the aspect is determined to provide a density variance.

13. The method of claim 1, wherein the determining density information comprises:

determining a centroid of intensities within the region identified as the aspect to provide a density centroid.

14. The method of claim 1, wherein the determining density information comprises:

regions of highest intensity within the regions identified as the aspect are determined to provide a most dense region.

15. The method of claim 1, wherein the characterizing the aspect comprises:

comparing the density value to a density value database to determine a condition associated with the tissue structure.

16. The method of claim 1, wherein the characterizing the aspect comprises:

comparing at least one of the thickness of the aspect, the size of the aspect, and the location of the aspect to a database to determine the relative condition of the tissue structure.

17. The method of claim 1, wherein the tissue structure comprises a blood vessel.

18. The method of claim 1, wherein the tissue structure comprises a carotid artery.

19. The method of claim 1, further comprising:

determining a total area of the aspect, wherein the characterizing the aspect further uses information about the total area.

20. The method of claim 1, further comprising:

determining a relative position of the aspect with respect to another aspect of the tissue structure, wherein the characterizing the aspect further uses information about the relative position.

21. A method for characterizing aspects of a tissue structure appearing in an image, the method comprising:

analyzing the image to identify the aspect of the tissue structure, wherein analyzing the image comprises: analyzing the columns of pixels to locate at least one of a lumen, intima, media, adventitia, a lumen/intima boundary, and a media/adventitia boundary;

determining density information for the aspect by applying an adjustment factor to intensity information associated with the image, the adjustment factor depending on a known density of one or more landmarks in the image, the one or more landmarks having a relatively constant density for different images; and

using said density information to characterize said aspect, said characterization providing information about the severity of a condition associated with said anatomical aspect, wherein a database comprising a plurality of plaque densities from a plurality and different populations associated with the occurrence of an event is used in conjunction with personal plaque density measurements for screening.

22. The method of claim 21, wherein the determining density information comprises:

determining an average intensity of the image within the region identified as the aspect so as to provide an average density.

23. The method of claim 21, wherein the determining density information comprises:

determining a peak intensity of the image within the region identified as the aspect to provide a peak density.

24. The method of claim 21, wherein the determining density information comprises:

determining a maximum intensity within the region identified as the aspect minus a minimum intensity within the region identified as the aspect to provide a density range.

25. The method of claim 21, wherein the determining density information comprises:

an intensity histogram is determined for intensities within the region identified as the aspect to provide an intensity histogram.

26. The method of claim 21, wherein the determining density information comprises:

a standard deviation of the intensity within the region identified as the aspect is determined to provide a density variance.

27. The method of claim 21, wherein the determining density information comprises:

a centroid of intensities within the region identified as the aspect is determined to provide a density centroid.

28. The method of claim 21, wherein the determining density information comprises:

the regions identified as the highest intensity within the aspect are determined to provide the most dense regions.

29. The method of claim 21, wherein the density information comprises information about a stiffness of the aspect of the tissue structure.

30. The method of claim 29, wherein the stiffness is associated with calcification of the aspect of the tissue structure.

31. The method of claim 21, wherein the characterizing the aspect comprises:

a database of similar aspects of the organizational structure is referenced.

32. The method of claim 21, wherein the characterizing the aspect comprises:

the historical information is referenced to provide said information about the severity of the condition.

33. The method of claim 21, wherein said analyzing said image comprises:

analyzing intensity gradients within the image.

34. The method of claim 21, wherein said analyzing said image comprises:

the data is identified with image intensity information.

35. The method of claim 21, wherein said analyzing said image comprises:

tissue boundaries are identified using image intensity information.

36. A system for characterizing aspects of a tissue structure appearing in an image, the system comprising:

a processor-based system receiving an input of a digital representation of the image, the processor-based system including circuitry operative to analyze the image by analyzing columns of pixels to locate at least one of lumen, intima, media, adventitia, lumen/intima boundaries, and media/adventitia boundaries and normalize the intensity information by applying an adjustment factor to intensity information of the digital representation of the image, the adjustment factor depending on a known density of a plurality of landmarks in the image, the plurality of landmarks having a relatively constant density for different images, the processor-based system further including circuitry operative to identify the aspect of the tissue structure in the digital representation of the image, the processor-based system further including circuitry operative to determine density information regarding the aspect of the tissue structure using the normalized intensity information, and the processor-based system further comprises circuitry operable to characterize the aspect of the tissue structure as a function of the density information, wherein a database containing a plurality of plaque densities from a plurality and different event-associated populations is used in conjunction with the individual plaque density measurements for screening.

37. The system of claim 36, further comprising:

a display that outputs the density information to a user.

38. The system of claim 36, further comprising:

a printer that outputs the density information as a hardcopy report.

39. The system of claim 36, wherein the aspect of the tissue structure comprises a region of atherosclerotic plaque.

40. The system of claim 39, wherein the atherosclerotic plaque is characterized as a function of stiffness.

41. The system of claim 36, wherein the tissue structure comprises a blood vessel.

42. The system of claim 36, wherein the tissue structure comprises a carotid artery.

43. The system of claim 36, wherein the density information indicates a stiffness of the aspect of the tissue structure.

44. The system of claim 36, wherein the characterization of the aspect of the tissue structure provides information about a severity of a condition associated with the aspect of the tissue structure.

45. The system of claim 36, wherein the characterization of the aspect of the tissue structure corresponds to a risk level associated with the aspect of the tissue structure.

46. The system of claim 36, further comprising:

a database of intensity information normalized for organizational structural aspects associated with a plurality of individuals.

47. The system of claim 46, wherein the database further comprises an event history, the event history associated with an aspect of the organizational structure.

48. The system of claim 46, wherein the intensity information normalized for an aspect of tissue structure is a normalized plaque value.

49. The system of claim 36, wherein the plurality of landmarks include adventitia as an upper tissue density limit and a lumen as a lower tissue density limit.