DE102006025917A1 - Object`s e.g. vascular tree of patient, outline determination method, involves determining brightness region between upper pixel values and lower pixel values representing objects e.g. vascular tree of patient, in histogram of x-ray image - Google Patents

Abstract

The method involves determining a brightness region between upper pixel values and lower pixel values representing objects e.g. vascular tree of a patient (4), in histogram of an x-ray image (2). A planar image is produced from the x-ray image, where pixels with pixel values in the brightness area are set to a value in the planar image. The planar image is straightened by an edge filter, where pixel in a transition area is assumed to be larger than an intermediate value. An outline is formed from the planar image as the quantity of image pixels formed with the intermediate value.

Description

The The invention relates to a method for determining the contour of a presented in a medical digital fluoroscopic image Object.

In Medical technology is increasingly adopting so-called roadmap procedures by. In this case, for example, after recording a DSA (digital Subtraction angiography) a so-called peak opacity image (peak Op) generated. This contains the Way one while the DSA in the represented object, e.g. the patient, wandering Contrast bolus or the brightest point (peak) of the bolus. For example, in a vascular treatment a patient is thus contrasted by the contrast agent Vascular tree in the Patients clearly visible.

During the Container handling becomes a catheter inserted into the vessel in x-ray added. The current X-ray image of the instantaneous position the catheter is in a further subtraction step with the Peak surgical image linked. in the resulting image are then the high-contrast catheter and the high-contrast vessel of the patient shown together. The vascular tree shown in the Peak Op image becomes so for the orientation of the vascular surgeon used in the treatment as a roadmap.

There Both the catheter and the contrasted vascular tree high contrast are, can in the sum-picture, when presented together, they hardly or only badly be differentiated from each other. To have an appropriate roadmap procedure at all carry out to be able to The operating mode of the X-ray system must not be too changed because the adjustment of the system, e.g. concerning iris diaphragm, by the mode change being affected. A switch from DSA to digital radiography or fluroscopy and the subsequent switch back to DSA supplies because of the hysteresis of the X-ray system Images that can not be used in the subtraction process. At worst, So another DSA must be done with further contrast agent. Such an X-ray mode switching may also accidentally be e.g. by pressing a wrong footswitch from the doctor, happen.

It is, however, e.g. from the German submitted on 27 December 2005 Patent Application 10 2005 062 445.6 known from the roadmap image the contour of the displayed object, e.g. the vascular tree, to extract and then only this then in the right place in any X-ray image, which then the catheter or the like represents, show. This creates a roadmap image, which is so the medical digital fluoroscopy image is, and in addition that object of interest, in the example the vascular tree, clearly visible.

task The present invention is the contour of the subject object in the fluoroscopic image.

The Task is solved by a method according to claim 1. The digital fluoroscopic image, e.g. a peak surgical image is off constructed of different image pixels, each pixel a specific Has (number) value. In the simplest case of a grayscale image these are gray values, e.g. between "0" and "255". The picture represents as a medical image always an object of interest, e.g. one Vascular tree one Patients, for The present method always applies that the object, ie its Image in the fluoroscopic image by its brightness (or color in the case of a colored picture), characteristic of the rest Picture takes off. For example, the object is very bright, where the rest of the picture is rather dark. The object thus has one characteristic brightness range from the rest of the image.

Around to use these differences in brightness, therefore becomes a histogram of the fluoroscopic image, e.g. in a diagram the greyscale values are plotted on the abscissa, the image pixels of the fluoroscopic image can accept (Dynamic range) and on the ordinate the respective number of pixels, which have the corresponding gray value.

The The invention is based on the following finding: Because of the brightness differences to the rest of the picture is in such a fluoroscopic image clearly and clearly recognize the object of interest or against the rest of the picture is demarcated. This is because the values of the Image pixels in the area of the object are clearly different from those of the rest Differentiate fluoroscopy image. In the histogram, that is the brightness distribution reflected in the picture, therefore, the object is clearly a particular Brightness range or object range can be assigned. The brightness range is usually an interval of pixel values that is on both sides of depending on a lower and upper pixel value is limited.

Therefore, according to the invention the object representing the object Brightness range between a lower and an upper pixel value determined in the histogram. Thus, the pixel values are known in the fluoroscopic image substantially correspond to the image of the object.

From the fluoroscopic image therefore a surface image is generated. In this, each image pixel ent ent picture pixel in the fluoroscopic image speaks, whose pixel value is in the brightness range, set to a first value. Image pixels with the first value in the area image thus correspond to the image of the object. All other image pixels of the area image, which are thus not assigned to the object in the fluoroscopic image, are set to a second value.

The such as area picture modified fluoroscopic image thus includes only pixels, either have the first or second value. The object is thus uniform, So "areal", in the area image shown. For the next Step is essential, that between first and second value at least an intermediate value exists. It can also be an interval of Intermediate values act.

All Places in the area image, to which a transition between pixel values of first and second value can thus be understood as object contour. Therefore, according to the invention the fluoroscopic image is smoothed with an edge filter. The Edge filter "wears away" by its smoothing property always the transitions between the different values of image pixels, thus always in the area of Object contour, where it produces the intermediate values as a transitional value from the first to the second value.

The Contour of the object is then called the set of all pixels in the fluoroscopic image which have the pixel value of the intermediate value (s).

The inventive method is particularly quick and effective feasible because of its simplicity. For edge extraction in the fluoroscopic image its accuracy is sufficient. At strong noise-laden images, the process is extremely reliable.

Of the in particular, the first value can be "2", the second value "0" and the intermediate value "1". This is how the area image can be get by with only two bits per image pixel. The process is characterized fast and needed little storage space.

There the object clearly differs from the rest with its brightness range Image, this or the brightness range is usually by a maximum or at least by a unique peak, ie a local maximum, represented in the histogram. The object area can therefore be determined in the histogram such that a maximum or a peak in the histogram surrounding area is selected. The area is here mostly through the flanks of the peak or a transition to a flatter Range clearly visible in the histogram. As in the picture viewing the coloring of the Object, e.g. light or dark, opposite the rest of the picture, known is, the peak in the histogram can be easily identified, e.g. near its lower or upper end.

As mentioned above, is the object in a fluoroscopic image representing the object histogram-sided characterized by a unique peak. Of the the area surrounding the peak is therefore unique with the corresponding one Object correlates and can therefore easily be used as a brightness range or Object area are determined. This can e.g. through a corresponding Observer of the histogram manually, computerized or by a fully automated process.

Of the Object area can also be so from the remaining fluoroscopic image be demarcated that first the absolute maximum of the histogram as the center of the object area, namely actual peak is determined. In the outside of this peak, so part of the histogram lying in the object area becomes a limit value for the Frequency of ascertained image pixels. This limit is chosen so that he just just left the rest of the image area without the one of interest Object characterized. Image pixels with pixel values whose frequency is above the threshold, so are counted to the object, the not others. In the method, therefore, the pixel values of all pixels with pixel values, their frequency in the histogram above the limit, all other pixels are at the first value set the second pixel value.

in the In the simplest case, the limit value is determined so that the limit criterion after the second highest looking for relative maximum in the histogram. As a limit is thus the maximum value of the frequency the maximum outside of the object area, that is usually below the peak and thus the second highest Maximum, chosen. This ensures that in the picture all actually not to the object area counting Brightness values, so pixel values are set. A relative Maximum is characterized by the fact that on both sides, so to bigger and smaller pixel values, the frequency falls in the histogram.

Of the above limit can be determined by an operator of the procedure be readjustable. Thus, by real-time image viewing e.g. personal Experience of the operator a correspondingly appropriate and meaningful Appearing area of the fluoroscopic image as an object area chosen and thus the correspondingly correct edges of the object be determined as a contour.

Before creating the histogram from the fluoroscopic image, this can be smoothed, eg with a corresponding smoothing filter. This leads to a suppression of the noise caused spikes in the histogram. The spikes could erroneously lead to the incorrect localization of the abovementioned maxima, and thus to the misinterpretation of the brightness (object) range in the histogram. Also, the highest and second highest maximum value in the histogram could be falsified. This would also lead to a wrong localization of the brightness (object) area in the histogram.

A about three to ten times smoothing has proven to be particularly successful in practice.

The inventive task will continue to be solved by a device according to claim 8. The device is designed such that it the inventive method To run can and in particular has a correspondingly formed Evaluation unit. The advantages of the device have already been explained in connection with the method according to the invention.

For another Description of the invention is directed to the embodiments of the drawings. They show, in each case in a schematic outline sketch:

1 a fluoroscopic image (mask image) of a patient before a contrast agent administration,

2 a fluoroscopic image (peak op picture) of the same patient after contrast administration,

3 a subtraction image of the images 1 and 2 .

4 a histogram of the subtraction image 3 .

5 one out of the picture 3 generated area image containing only pixel values "0" and "2",

6 the area image 5 after antialiasing,

7 a from the smoothed area image of 6 generated contour image,

8th a live image of the patient with catheter, with superimposed contour image 7 .

9 the overlay image 8th with additionally overlaid mask image 1 ,

1 shows an x-ray image 2 a patient 4 or its pelvic area. At the patient 4 a DSA examination with subsequent catheter treatment in the pelvic area is to be carried out. The X-ray picture 2 is therefore taken before the administration of a contrast agent as a mask image or native image. It only shows the bone structures 5 of the patient 4 , Blood vessels are not visible.

The patient 4 is now administered contrast agent, which is through its blood vessels or vessels 6 spreads in the flow direction of the blood. In the course of the propagation of the contrast agent or the contrast agent bolus (according to the X-ray image 2 ) continuously recorded X-ray images not shown. These X-ray images are summed in the context of the DSA in a known manner or from this a in 2 illustrated peak Op image 8th generated. In the peak Op picture 8th are next to the x-ray picture 2 illustrated bone structures 5 also contrasted with contrast medium blood vessels 6 to recognize.

As part of the DSA is now the peak Op image 8th the x-ray picture 2 subtracted to become one in 3 shown subtraction image 10 to get. Peak Op Image 8th and X-ray image 2 Although differ essentially by the illustrated blood vessel 6 , but are also related to the bone structure shown 5 of the patient 4 not absolutely identical. By subtraction, therefore, not all of them are next to the blood vessel 6 shown body or bone structures 5 , in the example the bones 5 of the pelvic area of the patient 4 completely extinguished, but it remains artifacts 12 in the subtraction picture 10 ,

In contrast to the strong or dark or almost black contrasted blood vessel 6 But these are the artifacts 12 rather gray. The remaining image area of the subtraction image 10 is very bright, almost white.

In the subtraction picture 10 should now as exactly as possible the outline 13 or the edge or contour of the blood vessel 6 be isolated. The subtraction picture 10 (and usually the other X-ray images) is in digital form or can be digitized accordingly, so that it from individual image pixels 14 is built, which in 3 in the upper left corner of the subtraction image 10 are indicated. The subtraction picture 10 is an 8-bit grayscale image, ie each image pixel 14 has a brightness value between "0" and "255". The value "0" corresponds to white, the value "255" to black.

From the subtraction image 10 Now a histogram is made, ie the image pixels 14 , which have a certain gray value, are counted and their number is plotted as a frequency H in a graph above the respective gray value P. The histogram 16 of the subtraction image 10 is in 4 shown.

Because the blood vessel 6 in the subtraction picture 10 is shown particularly dark, have this representing image pixels 14 correspondingly large gray values P. The peak 18 in the histogram 16 therefore corresponds to the image pixels 14 that together the blood vessel 6 in the subtraction picture 10 represent, or the entire area 20 the dark to black image pixels 14 , The artifacts 12 on the other hand, the subtraction image is more in the middle gray, which is why in the histogram you are roughly the area 22 correspond. The remaining bright area in the subtraction image 10 roughly corresponds to the area 24 in the histogram 16 with very small gray values.

In a first variant of the method according to the invention, the peak 18 and with this the blood vessel 6 corresponding area 20 of gray values by subjective observation, for example by a human observer, not shown, or an automatic analysis of the histogram (eg maximum search, curve discussion, other methods of mathematics, pattern recognition or the like) determined. So, the lower limit 26 and upper limit 28 of the area 20 determined with respect to the gray values P. As lower limit 26 For example, a value of approximately "180" results in the upper limit of the value "255".

The The following numerical values refer to these by way of example mentioned values.

A second embodiment of the method is in 4 shown. Here is next to the peak 18 belonging to the absolute maximum H1 of the frequency H, the next lower relative maximum H2 in the histogram 16 certainly. This is located in 4 approximately at a gray value P of "20" and represents the entire bright image area in the subtraction image 10 which are not the blood vessel 6 or artifacts 12 represents. Finding the maximum H2 in the histogram 16 thus represents a threshold criterion to determine a histogram threshold, namely H2, which is precisely the blood vessel 6 from the remaining image areas of the subtraction image 10 demarcates.

Based on the limit value H2 now become the lower limit 26 and the upper limit 28 established. All image pixels 14 in the subtraction picture 10 , whose brightness value P occurs with a frequency of less than H2, do not become the object, ie the blood vessel 6 expected. Therefore, the lower limit 26 set to the value P = 180. For P <180, the frequency is the image pixels 14 namely smaller than H2. The object area in the histogram 16 So it starts at the lower limit 26 at P = 180. The image pixels with brightness values between P = 180 and P = 230 have a frequency greater than H2 and are thus counted to the object area. From P> 230, the frequency of image pixels decreases 14 again under H2. The upper limit 28 is thus determined to be P = 230.

Now all image pixels are out of range again 20 , ie for P <180 and P> 230, set to the brightness value P = 0. All other image pixels with original brightness values between 180 and 230 are set to the uniform value P = 2 and thus corresponding again essentially to the region of the blood vessel 6 ,

The corresponding image pixels with values "0" and "2" become in one area image 29 set that in 5 is shown. This is a 2-bit grayscale image with pixel values between 0 and 2. The blood vessel 6 is therefore no longer in different brightness shades, as in 1 to 3 , but only uniformly colored.

The area picture 29 out 5 is now smoothed with a suitable edge filter. In the example this is eg a 3 × 3 binomial filter with a core of 121,242,121. A corresponding edge filter "worsens" the respective transitions between the areas P = 0 and P = 2 in the area image 29 which are just the edges 13 of the blood vessel 6 correspond. At just these edges 13 arise through the binomial filter image pixel 14 with values P = 1, in 6 represented by a the blood vessel 6 surrounding edge area 30 , The smoothing by the filter is tiring precisely the steep transitions between pixel values "0-2" to smoother, smoothed transitions "0-1-2".

Because these transitions are exactly the outline 13 of the blood vessel 6 in the area image 29 match, all image pixels become 14 with the value P = 1 from the area image 29 searched or extracted and according to 7 in an edge picture 32 saved. Any missing gaps in the border area 30 are closed, for example, by its interpolation. The resulting edge area 30 thus corresponds to the outline 13 of the blood vessel 6 and is now in the road process in an isolated edge image 32 available for further processing.

8th shows a first application in the further X-ray images 34 of the patient 4 were generated. From the measured native image, the mask image is subtracted, so that the bone structure is no longer recognizable. However, a radiopaque high-contrast catheter is recognizable 36 , the doctor in the blood vessel 6 the patient is introduced. The blood vessel 6 is in the x-ray 34 also not recognizable, since it no longer contains contrast media. For orientation for the doctor is in addition to the X-ray 34 the edge picture 32 or the filled vessel displayed in the right place to the doctor by showing the outline 13 of the blood vessel 6 , so the Randbe Reich 30 To give guidance where the catheter is 36 in the blood vessel 6 currently located.

9 shows an alternative representation of an X-ray image 34 , Here is not just the catheter 36 visible, because next to the insertion of the border area 30 from the edge picture 32 In addition, the measured native image, so the X-ray image 2 out 1 , is displayed correctly. This shows the doctor not shown next to the edge 30 of the blood vessel 6 for better orientation also the skeleton structure 5 of the patient 4 , Thus, the doctor has more information as to exactly where the catheter is 36 in the patient 4 currently located.

Because the edge picture 34 isolated, so no artifacts 12 contains more, it can also be used for other subtraction methods since there is no non-subtractable overhead image information except the edge area 30 is included.

Claims (8)

Method for determining the contour ( 30 ) in a medical digital fluoroscopic image ( 8th ) ( 6 ), where the object ( 6 ) in the picture ( 8th ) a characteristic brightness range ( 20 ) compared to the rest of the picture ( 22 . 24 ), comprising the following steps: a histogram ( 16 ) the frequency (H) of the image pixels ( 14 ) of the fluoroscopic image ( 8th ) is created via their pixel values (P), - the object ( 6 ) representing brightness range ( 20 ) between a lower ( 26 ) and an upper ( 28 ) Pixel value in the histogram ( 16 ) is determined, - from the fluoroscopic image ( 8th ), a surface image ( 29 ), in which all pixels ( 14 ) with pixel values (P) in the brightness range ( 20 ) to a first value ("2"), all others to a second ("0") value, with at least one intermediate value ("1") existing between the first ("2") and second ("0") values , - the area image ( 27 ) is smoothed with an edge filter, where pixels ( 14 ) in the transition area ( 30 ) assume between the first ("2") and second ("0") value at least for the most part the intermediate value ("1"), - the contour ( 30 ) is taken from the area image ( 29 ) as the set of all image pixels ( 14 ) is formed with the intermediate value ("1"). The method of claim 1, wherein the first value is "2", the second value is "0" and the intermediate value is "1". Method according to claim 1 or 2, wherein the brightness range ( 20 ) the maximum (H ₁ ) of the histogram ( 16 ), in which the brightness range ( 20 ) is determined such that an area surrounding the maximum (H ₁ ) in the histogram is selected. Method according to one of the preceding claims, wherein the brightness range ( 20 ) the maximum (H ₁ ) of the histogram ( 16 ), in which: - in the histogram ( 16 ) the second highest maximum value (Hz) of the frequency (H) is determined as the limit value, - the brightness range ( 20 ) is determined such that it contains all the pixel values (P) whose frequencies (H) in the histogram ( 16 ) are above the limit. Method according to claim 4, wherein the limit value can be readjusted by an operator of the method. Method according to one of the preceding claims, in which the fluoroscopic image ( 8th ) before creating the histogram ( 16 ) is smoothed. The method of claim 6, wherein a three to ten times smoothing he follows. Device for determining the contour of a in a medical digital fluoroscopic image / peak surgical image shown Object, wherein the object in the image has a characteristic brightness range across from the rest of the picture has, with A histogram unit for Create a histogram of the frequency of image pixels of the Transillumination image over their pixel values, and to determine an object representing the object Brightness range between a lower and an upper pixel value in the histogram, - one Evaluation unit for generating a surface image from the fluoroscopic image, in which all pixels with pixel values in the brightness range to one first value, all others are set to a second value, wherein between the first and second value at least one intermediate value exists, - one Edge filter for smoothing of the area image, where Pixels in the transition area between the first and second value, at least for the most part, the intermediate value accept, - one Evaluation unit for forming the contour from the area image as the set of all image pixels with the intermediate value.