GB2507358A - Evaluation of imaging quality of medical diagnostic ultrasound scanners based upon void counting in a 3D volume using voids or low echoic cysts - Google Patents

Abstract

Ultrasound quality assurance testing is disclosed comprising: 3D imaging a phantom comprising tissue mimicking material and anechoic void features; calculating void detectability ratios (VDR) for every point in the volume; and counting the number of voids for each depth for a number of specified minimum contrast or VDR levels to provide quality assurance values (see graph fig. 11). An aspect of ultrasound imaging quality is the capability of a scanner to detect small lesions, presenting as low echoic structures. This invention uses the number of random voids detected from a 3D-volume scan, of a tissue mimicking material, to provide evaluation criteria. The volume data is used for calculating the void detectability ratio (VDR). A higher density of detected random voids indicates higher quality. Any VDR-level exceeding the void detectability limit indicates a higher resolution. This is because higher counts are effected by smaller voids being detected, down to the resolution limit of the scanner. By calculating different fitted curves for the resulting plots and evaluating the aberration of the fitted curve from the measured void count additional quality information is provided.

Description

Evaluation of imaging quality of medical diagnostic ultrasound scanners based upon void counting in a 3D volume using voids or low echoic cysts.

Background

This invention relates to a device to deternunc the imaging quality of diagnostic ultrasound scanners using the number of voids detected to provide evaluation criteria..

Ultrasound images in medical diagnostics are provided by ultrasound scanners working on pulse-echo principle. These machines emit ultrasonic pulses into the tissue and use the amplitude of the back-reflected signal as a gray or color level depicted on the screen along a line corresponding to the path of propagation within the body. By placing a number of these scan lines, side by side, a 2D image plane is created as shown in Fig 3 (B-Image top left). By storing several adjacent images complete scan volumes may be stored (See Fig 3 B, D and C image and Fig 4 top left). One aspect of overall imaging quality,is the capability of a scanner to detect small lesions which may present as low echoic structures. The discrimination of low echoic structures (or voids) is more difficult to achieve than showing small high echoic structures. Determining the void detection capability of such systems is important for several reasons: * Comparing scanners before purchase.

Determining if a scanner is adequate for certain diagnostic purposes o Continued quality assurance during scanner lifetime o Evaluahng an ultrasound scanner following a repair A number of quality assurance processes have been used so far, ranging from simple 2D evaluation of a single image to complex evaluation of 3D-volume scans, which are generated by moving the 2D probe over the test object, using the volume data for calculating void detectability ratios (VDR) . The VDR calculation is the baseline for the current invention.

Any 2D solution is inadequate to test the geometry of the three-dimensional acoustic pulses used.

To provide repeatable patient-independent measurements all these approaches use tissue-mimicking material phantoms (TMM-phantoms). To measure voids, and to be able to compare values measured against a baseline, an a-priory knowledge of the void's geometry, size and orientation was originally deemed useful (e.g. using spherical or short cylindrical voids with given diameters. This restricts the comparison of resolution to the void sizes offered and often restricts the scanning geometries, that can be tested (Linear arrays are fully covered in most cases).

Another approach is the use of a random void phantom, however it is difficult to obtain useful resolution information from the contrast or VDR-levels. as the void distribution is random and there is no luowledge as to the real size of a specific void.

A visual inspection of the reconstructed transparent 3D-volume image showing the random voids as translucent objects gives a visual impression of imaging quality and how it changes with depth. A higher density of detected random voids indicates higher quality, but does not provide any quantifiable results.

Statement of invention

The present invention is based on the evaluation of 3D-scans of TMM-phantoms and helps to overcome the lack of quantifiable results from random void detection, by counting the number of voids detected per depth and VDR range. (i.e. the VDRv, which is the maximum VDR-level found within the specific void, determines at which level the void is counted).

For any VDR-level exceeding the void detectability limit, (i.e. a VDR of approximately 2.5) higher counts indicate a higher resolution. This is because higher counts are effected by smaller voids being detected, not by noise. The void-count thus offers quantitative infomation for random void phantoms, which practically cover a continuous range of void sizes a maximum void diameter found in the 1MM down to the resolution limit of the scanner. This simple approach however does nol offer absolute values for the void-sizes detected.

By calculating different fitted curves for the resulting plots or parts of these plots and evaluating the aberration of the fitted curve from the measured void count additional quality information is provided.

Using phantoms with a well defined diameters and number of voids may provide the calibration needed to correlate counts at a certain VDR level to the absolute void sizes detected in the random void phantom.

Advantages This method may be used both for woricing with tissue-mimicking material (TMM) phantoms, using well defined voids, as vell as with 1MM-phantoms using randomly distributed voids.

The method allows the evaluation of data from a random void phantom and so it is possible to extend the quality control from linear array scanners to curved-and sector-array-scanners.

For higher \DR-levels (i.e VDR-levels larger than 2.5)VDR values at a given depth from the ultrasound transducer correlate with the relative void sizes. This allows generating curves that give a correlation bePveen relative void size and void count at a given depth.

Using the void-counting procedure on both standard-void and random-void fhantoins makes it possible to calibrate the relative void size derived from VDR-values of random void measurements to real void sizes.

By extending the void-count down to VDR-levels below the void detection range (i.e VDR-levels smaller than 2.5), where additional count will stein predominantly from noise, it is possible to evaluate noise based contributions to overall count, improving the relevance of the measurement in the higher VDR-regions.

Noise in the case of ultrasound imaging, contains both time-invariant acoustic speckle-pattern and the electronic noise. In the focal zone, ( see Fig 1 between near and far field) where a clear speckle pattern is to be expected, the void-count versus VDR curve extends from the region of high VDRs down to the noise region of the VDR in a way that a power-law fit will provide excellent correlation. Other regions cannot be fitted as well. The quality of fit can he used as a quality measure for imaging.

Introduction to drawings

Drawings are included, following the detailed description, to help provide a clearer understanding for the reader of this invention. A summary of each drawing is provided below. Fig 1

A diagram of the ultrasound beam passing through a cylindrical void Fig2 Describes the image plains B, D anti C and the software presentation of cylindrical voids shown inFig 1 Fig3 Ultrasonic representation of a Random Void Phantom in standard 3D presentation Fig 4 3D reconstructed presentation of the greylevels of the random void mteridl and of the reconstructed VDR magnitude of each voxel of the random void material shown in Fig 3. Fig 5

Illustration of voids, counted from the skin line, with \TDR level windowed between 255 and normalized levels of VDR. Fig6

Connection of a computer to a ultrasound scanner. Fig7

Illustration of the differentiation between points belonging to voids and points belonging to TMM. Fig 8

Illustration of how the void counting software fits sphere approximations into the grey level chart. Fig 9

Shows for Pig 8, the normalized VDR amplitude \vhich provides the brighrness value of each approximated sphere.

Fig 10 Logarithmic plot of power curves.

Fig Ii Evaluation of software to count voids using cylindrical void.

Addendum Terms and definitions used.

Detailed description.

Before we can understand the reason for void counting it is important to understand how to calculate the Void Detectability Ratio (VDR).

Equipment used The equipment used required for die VDR measurement is shown in Figure 6. It consists of a phantom, transducer positioning device with platfonn, PC including software for image recording (i.e. frame grabber software) and analysis and hardware for making connections to thcvideo output of the diagnostic ultrasound system (i.e. frame grabber or digitizer hardware) and the positioning device. This arrangement is suitable for acquiring 3D ultrasound image data using 2D ultrasound imaging systems.

At the top of the phantom the transducer is held petendieular to the B-sean plane and in one dimension, the transducer slides over the phantom surface via a coupling medium.

2D-iinages from the diagnostic ultrasound system are transferred in real time to the PC by applying a digitizer or converter, depending on the available signal output fimetion of the diagnostic ultrasound system. Between each of the images the transducer is moved by a certain distance in a direction perpendicular to the scan plane. In most eases a sensor for probe displacement provides a signal used to accurately reconstruct, in 3D, the image frames.

In some cases the displacement is calculated from the velocity of displacement The acquired three-dimensional data are stored in a matrix and can be displayed in three different orthogonal planes. The B-plane is die scan-plane of the transducer. The two orthogonal planes are the C-plane parallel to surface and the D-plain is orthogonal to both B-and C-plane. (see Fig.2 and Fig.3 for examples of cylindrical and random voids in their respective image planes.) Imaging voids When imaging voids the backscatter from the tissue or tissue mimicking material, which is a speckle pattern due to interference of the different parts of die ultrasonic beam, provides the general background. Voids have no reflecting structures inside, thus for an ideal image should have no signal. Figure 1. shows the ultrasound beam passing through a artificial cyst or "void", within the phantom. A wider beam or smaller diameter cylinder means part of the ultrasound beam no longer passes through die cylinder and will produce backscattcr from the TMM, winch seems to originate from die void. The void will no longer seem nonechoic.

The useftml signal for detecting die void is the mean signal level of the background reduced by the iuean residual signal at die void position. This signal must be visible above the variation of the background level (i.e. die variation due to the speckle pattern and electronic noise generated by the system).

For images with a varying background an objects detectability is generally described by the signal to noise ratio (SNR) given by the ratio of signal over the root mean square (RMS) of the background variation, The smaller the SNR, the greater the dangers of losing tile signal in the background or mistaking background variation for a void signal.

The voids examined are so small, that the signal at the edge of the void the amplitude levels do not give a sharp drop, but continuously reduce towards the center of the void. Therefore it does not make sense to calculate a mean void value Thcrcfore for ulh-asound imaging SNR has heeti replaced by a. value called void detectability ratio (VDR) to define detectability of the void signal.

Calculation of VDR Please refer to addendum 1 for definition of terms Ltsed To calculate values of VDR the following procedure is adopted. Within a volume selected calculations are perthnned for each C-plane..For the explanation of the procure however we use the data along a single line as shown in Fig 7.

The automated discrimination between regions corresponding to voids and regions belonging to foam acting as a tissue mimicking material (TMM) is made with the following algorithm: For each C-plane an average signal level p3 is calculated. All points with a grey level gi <113 are classified as void, the remaining points are classified as TMM. High gj values due to specular reflection at the bottom or top of voids are included in die TM1VI-values. For all the points of TM.M the mean value ti and the standard deviation ci are calculated. Then\TDR is calculated hr each voxel in the image uing its respective grey level g VDRiu-gj)/c1 1 is not only calculated for voxeis, which had been eiacifled as voids but ai9o/or tho.ce, that lie wi hui the TMM-regi on The largest value of VDR occurring in a particular region (a line, a plane, a ROT or a void) is used to characterize the signals in that region; of course this corresponds to die point with the lowest grey level.

Thus \TDR for a void is calculated from the lowest grey level g\I occurring inside that void.

Following VDR measurement and rendering of the 3D data as VDR levels for each voxcl in the ROT and are stored in a 3D matrix Again this can be displayed in B-, C-and D-images as shown in Fig 2. The greatcr the \TDR, the brighter the voxel is displayed.

The centre chart in this figure, is a representation of the maximum 3D VDR levels detected in each of the C-planes. The chart shows that the level drops as the depth increases This is due to the ulü-asound beam widening in front of and behind the focus. As the beam widens is no longer completely fits through the void as shown in Fig 1. The outer part of the beam, which hits the TMM, is back-reflected thus reducing the modulation depth of the void.

Using the Random Void Phantom (RVP) Fig 3 clearly shows an ultrasound image of the RYP foam material. Ft is displayed in three orthogonal image planes B,C and D. These images clearly show that the voids of different sizes and shapes are distributed uniformly and randomly throughout the phantom.

Looking at the RVP, 3D reconstructed transparent image of the foam (Fig 4 LH image), one gets a good impression of the higher number of voids detected in the centrat region ie in the focal zone. Calculating VDR and looking at the 3D reconstruction (Fig 4 RH image), we now get a dense cluster at the focal point which again shows that the density of voids may be a good indication of vbid detectability.

This gave the idea that we could quantify our understanding by including a void count. This is the nub of the invention. When introducing void count it turned out that in addition to providing a simple count and density infonnation, this count when related to VDR levels, showed a relationship between imaging quality and quality of a Power Law fit.

There are several methods of perfonning void counts. A simple and fast procedure is described below. The vbid counting procedure is just an example of possible void counting concepts. Other concepts of counting exist such as segmentation and skelletation as they are used in topology. They will provide different quality in counts however they need very high processing Umes: Procedure for Void Counting: This is a simple and fast procedure which uses the fact, that voids of similar diameter to the beam-width will produce a conical drop of gray-levels towards the center of this void. This is shown in Fig 8 in the cross sections. This drop in grey level results in a clear peak in VDR as shown in Fig 9. This procedure stores the position and VDR of these peaks. Counting the number of stored peaks gives the number of voids. This works well for well focused scanners in the focal plane or around the detection maximum as peaks are well defined there.

Looking at the counting procedure results, shown in Fig 5, we find that the VDR level,( as shown on the central 2D chart in Fig 2. and the change in level with depth and focus) does not filly describe the vivid impression of high void detection the clustering of voids at the focus offer in the transparent 3D-image shown in Fig 4.

Showing information of both VDR and void density in one diagram will support the correct interpretation of data. Now void count is not identica] with void density as shown in Fig.4. To get to density information we need to have information of both the amount and the size of the voids. The procedure described above stores both the position and the amplitude of the VDR for each local maximum.

VDR-levels are a good indicator for the void size as long as their diameters are about the size of the beam width, This becomes evident when looking at the two lower right hand diagrams in FigS: As the beam moves from the one edge of the void towards its center the baclcscattered intensity reduces. If the void is not too large the signal will not reach its minimunt gray level (i.e. background noise level) but will pick up again as we get nearer to the second edge. The smaller the void, the less pronounced this dip in gray-level will he. This dip in gray-level is our "signal" and its amplitude is thus related to the VDR-level. Large voids thus have higher VDR than small voids. VDR can thus be used as a relative measure of void size at a given distance fi-orn the transducer. This is shown in FigS, where such spheres are placed above the gray-level cross-section to demonstrate this correlation.

A presentation of the voids. .so counted, is provided in FigS. This figure is similar to the central VDR-chart of Fig 2 but shows on the right hand side, an estimation of void count for a given depth (in this case 1 to 7 cm), and a total couffi. The overall maximum VDR-levels for each C-plane in from Fig.2 have been replaced by the maximum levels detected for each void separately (see VDRv). Each of the levels has a circle corresponding to its relative size printed at the VDR-level. By this Method the impression of the density of the voids thus corresponds well with the reconstructed 3D-image in Fig.4 and additionally has the void counts as measurement value for comparing the void detection of different systems.

Calibrating the VDR-size relationship The VDR-level in itself can only provide a relative size relationship at a given depth. To improve the density information an absolute value would be of interest. This can he derived by using a standard phantom with known void diameters. When evaluating these phantoms there will be a definite step in the VDRv dependant countindieating the detection of these voids. The VDR-value at this step gives the VDR-Value correlating with the void size. This can then be used to calibrate the void size detected for each VDR-level.

New procedures using void count versus VDRV-levels Counting the low echoic voids at a given depth, windowed for all VDR levels, adds new information hitherto undiscovered.

A relationship in the ultrasound detection of voids has been discovered between the number of voids for a given void size and VDR in a fixed volume. The mathematical relationship is a power law curve.

This power Law holds even when extending tile void-count down to VDR-levels below the void detection range (i.e VDR-ievcls smaller than 2.5), where additional count will stern predommantly from noise Noise in the case of ultrasound imaging, contains both time-invariant acoustic speckle-pattern and the electronic noise. In the theal zone, ( see Fig 10 between near and far field) where a clear speckle pattern is to be expected the void-count versus VDR curve extends from the region of high VDRs down to the noise region of the VDR in a way that a power-law fit will provide excellent correlation. This correlation is best at the point of maximum deteetahility.( maximum VDR on the VDR versus depth plot shown in Figure 2.) The fit will not be as good when a scanner shows poor void detection. When a scanner has high side lobes ie poor imaging quality the fit is poor even at the point of maximum detectability.

For C-Plane depths in (he near field and far nerd, away from the point of max mum detectability, the log curve is no longer straight therefoje the Power Law is no longer applicable.

Extrapolation of the noise-based part of the void-count to the region of higher V.DR-lcvels For the plots in fig 11, derived from real measurement data, the void-count vs. VDR-level for 2cm depth, which is near the focal point, and Scm depth, which is towards the end of the The data used lbr calculating the noise (+interference relate are marked by triangles and squarcs. An exponential curve is fitted to these regions and these fitted trend-lines are extrapolated to higher VDR-levels. Subtracting die extrapolated value from the measured value for a given VDR will provide a reduced void-count giving a value corrected for noise for the region, where the count of real voids predominates. The reduction in additional counts down to small values indicates the region where smaller voids can no longer be detected or discriminated from speckle "noise". One cannot give an exact limit of resolution, hut determining the crossover-point an the maximum difference between extrapolated noise and real count will provide an additional quality parameter.

Addendum.

Terms a ad Deflnitions tiy (@2:PELflZ jfl medium, usually fluid or a gel, that allows echo-free coupling of the transducer to the coupling window of the phantom.

ac:ysstilcTvc. in.rt 1et.icv arithmetic mean of the frequency f and f2 at which the amplitude of the spectrum of the acoustic signal, i.e. the output of a hydrophone placed in an acoustic field at a specified position, firsi becomes 3dB lower than the peak amplitude [Conforms to 3.2 of IEC 62306, ed. 1.0(2006-03):3.4.2 of IEC 61102, modified] arttLrtctua ch average grey-level value at a specific region in an image, where no signals are expected (e.g. inside the image of a void) attenuation coefficient at a specified frequency, the fractional decrease in plane wave amphtude per unit path length in the medium, specified for one-way propagation Units: rn' [attenuation coefficient is expressed in dR-rn by multiplying the fractional decreasc by 8,686 dB] - [3.3 of IEC 61391-2].

backscatter coefficient at a specified frequency, the mean acoustic power scattered by a specified object in the 1 800 direction with respect to the direction of the incident beam, per unit solid angle per unit volume, divided by the incident beam intensity, the mean power being obtained from different spatial realizations of the scattering volume Units: m-1 steradian-1 [cm-i steradian-1] NOTE The frequency dependency should be addressed at places where backscatter coefflcient is used, if frequency influences results significantly.

[3.6 of IEC 61391-i and 3.7 of lEO 61391-21 backscatter contrast ratio between the backscatter coefficients of two objects Of regions NOTE Dackscatter contrast can be frequency-dependent but it is independent of any image system.] [3.3 ofIEC 61 391-2]

C-

basic cross sectional presentations of 3D-images: B-image is in a plane that is created by the ultrasound scan-fines (scan plane) C-image is in a plane perpendicular to the central scan line in the B-image D-image is in a plane perpendicular to B-image-plane and C-image-plane B-,C-D-planes B-plane; scan plane C-plane; reconstructed image plane that is perpendicular to the central scan line in the scan plane image D-plane: reconstructed image plane that is perpendicular to the scan plane and the C-plane cine image loop (or cine loop) set of multiple B-mode images acquired sequentially in time or reconstructed to represent a time sequence, and available for viewing and processing either continuously or as individual images tytti,T iiTi' 4i: portion of the phantom's enclosure provided for entrance and exit of the transmitted ultrasound pulses to/from the TMM without significant attenuation or distortion NOTE: The coupling window usualiy consists of a thin membrane which protects the TMM frcm evaporation, leakage and mechanical damage by the transducer and which does nct significantly alter the ultrasound signals ItT ier: t o H smallest true value of the measurement, which is detectable by the measuring method [110 of IEC 60761-1. ed. 2.0 (2002-01)] [WC Glossary for SC45B] digitized image data: two-dimensional set of pixel values derived from the analogue or digital grey level values, as transmitted to the viewing screen of the ultrasound imaging device, containing the grey level information derived from the ultrasound echo signal. 12] grey-level value value of the digitized image data at a particular pixel in the image Note: The grey level values determine the brightness of specific pixels in the image and they usually range from 0 (black) to 255 (white).

(acoustic) scan line one of the component lines which form a B-mode Image on an ultrasound systei4i monitor, where each line Is the envelope-detected A-scan line In which the echo amplitudes are converted to brightness values [3.26 of W,C 61391-1 and 3.2 of IEC 61391-2] scan plane acquired Image plane containing the acoustic scan lines [3.30 of 61391-2] specific attenuation coefficient at a specified frequency, the slope of attenuation coefficient plotted against frequency [3.4 of 61391-2] tIssue mimicking material (1MM) material in which the propagation veioclty (speed of sound), reflection, scattering, and attenuation properties are similar to those of soft tissue for ultrasound in the frequency range I Milztol5MHz (3.36 of 61391-1 and 3.36 of 61 391-2( TMM 3D artifidal anechoic-cyst phantom Phantom containing tissue-mimicking material having a known speed of sound and attenuation coefficient, in which there are well-defined regions whose backscatter contrast is at least -60 dB

relative to the background

void; artificial anechoic cyst (synonym) region of defined geometry in a tissue-mimicking material that generates no scattered acoustic waves NOTE: Saline solution is known to produce extremely low levels of scattered signals and therefore it is an optimum approximation to a perfect void.

DetectabiLity The state of being detectabie void-detectability ratio (VFMt) number characterizing the visibility of an Image area corresponding of a void of defined diameter surrounded by 1MM in the phantom. The Image of the surrounding TMM material is expected to show modulated grey levels (i.e. an ultrasound speckle image) due to the uitrasound Interference patterns.

Note 1: The image of the surrounding TMM material is expected to show modulated grey levels (I.e. an ultrasound speckle image) due to the ultrasound interference patterns.

Note 2: Both maximum and mean values have been used to define a void detectability ratio deteetabilty ratio for a single voxel Detectability ratio for a single voxel is defined by: VDR = where = an.al.ne o..ua d.ig.i.ti: red. in.: age do a fm in a aid on j o..st c:s life t.ii a m.age or a

VOO

= i-the grey level value contained in the image region of a void cr1 = standard deviation of the digitized image data from a region just outside the image of a void Note 1: VDR formula is derived from reference [4] Note 2: For further details see A.3.2-A.3.4 maximum \PR Maximum VDR is defined by VDR = (p1-g) I G = Max=19 (VDR) where Va:ue OT rae di}tizrr. !mnaric. data from a realon f*u.st outside the i in.age ot. a 9min = maximum grey-level value inside a void i.e. the maximum difference between the average grey-level value of the surrounding TMM and the grey level values depicted in the void on = standard deviation of the digitized image data from a region just outside the image of a void Mean VDR Mean VDR is defined by: VDR = (p1 -L12) / a1 = 1/n(E1=j(kDR1)) where mean. vane a di =1 frii dv ad lm.aare do to tront * a net 0 atsld..e the tim. age ol a void iL.z)JC3fl \/a.Lttit Oi iit duitirtlO].iiikO. Onto *i jgj tIOO tea: agoa roe rce;eetlng o vo C cr = standard deviation of the digitized image data from a region just outside the image of a void VDR limit minimum value of the VDR for which there is unambiguous visualization of a void on an ultrasound image NOTE: En Reference [4] the detection limit for the detection of voids *of the defined void sizes (See 7.1) for a noise level independent of electronic noise was stated to be around VOR = 2,5 Signals above this threshold should he usable for diagnosis. The method is independent of the threshoEd VDR-value actually used, Symbols o = speed of sound (m s) g..= i-the grey level value contained in the image region of a void gmin = inininmrn grey-level value inside a void T= temperature (°C) S= salinity (parts per thousand: %) Sgmax = maximum value of the digitized image data (grey-level value) z= depth (m) VDR = void detectability ratio --averaged value over the image of a void VDR = void detectability ratio --maximum value over the image of a void VDR: = detectability ratio for a single voxel (pixel) i Mi = mean value of the digitized image data from a region just outside the image of a void Ms = mean value of the digitized image data from within the image area representing a void = standard deviation of the digitized image data from a region just outside the image of a void = standard deviation of artifactual echoes

Claims (5)

1. Claims I. A medical imaging ultrasound quality assurance device using tissue-mimicking phantoms and a 3D data-acquisition-setup to acquire 3D-volume data and calculating void detectability ratios (VDR) or signal to noise ratios (SYR) for every point of the volLLme comprising a module that performs a void-count for each depth for a number of specified minimum contrast-, or VDR-levels to provide quality assurance values.
2. 2. A medical imaging ultrasound quality assurance device according to claim 1, that extends void-counts below the usual detection limits for VDR-levels of 2.5 thus aLso counting in a region, where noise predominates.
3. 3. A medical imaging ultrasound quality assurance device according to claim 1 and claim 2, which provides ultrasound imaging quality information based on a power-law fit for the total range of low to high VDR-values for the void-count vs. VDR data plot.
4. 4. A medical imaging ultrasound quality assurance device according to claim 1 and claim 2, which uses an extrapolation of void-count versus VDR-values of the region with predominant noise to provide more reliable count values for VDR-leve!s above 2.5
5. 5. A medical imaging ultrasound quality assurance device according to claim 1, which provides means to combine the data of a random-void phantom and a phantom with fixed void diameters to allow calibration of the random void VDR-levels with respect to void size.

   6, A medical imaging ultrasound quality assurance device according to claim 1, which uses the void-count and void size to determine the void-density in a given volume.