Aaron Moskalik

Registration of three-dimensional compound ultrasound scans of the breast for refraction and motion correction

Use of multiple look directions, that is, compound imaging, has been shown previously to increase detection of specular reflectors and averaging of speckle noise in gray-scale images, often at the expense of spatial resolution and other misregistration errors. In color flow imaging, additional view angles can fill in vessels missed due to Doppler angle dropout and increase quantitative and visual Doppler accuracy by triangulation or a simple peak-frequency-shift combination algorithm. Image registration and unwarping throughout multiple three-dimensional (3D) volume sets should correct for many refraction artifacts, motion between and during compounded image sets and even, possibly, positioning errors between image sets, acquired months apart, to display growth of abnormalities. The registration described here does not provide sufficient accuracy for formation of enhanced coherent apertures, but shows promise in some cases to provide superior compound images and possibly comparisons of current and prior studies. In this study, the breast is stabilized by mild compression between a flat plate and a scanning membrane. Registration and unwarping is performed retrospectively on two separate volumetric data sets by defining pairs of corresponding points and, in some cases, line and plane segments. Three-dimensional linear affine transforms are performed using identified points, lines and planes. 3D nonlinear warped transforms are also possible given adequate numbers of identifiable points. More than two data sets are registered by selecting one as the standard, and registering the remainder to match. The most appropriate algorithm, such as averaging or maximum amplitude, may be used to combine the data sets for display. Significant success has been achieved in compound display of a test object and of the breast in vivo, even when there was relative motion or warping between image sets. In pulse-echo imaging, homologous feature registration for compounding appears to have advantages over mechanically registered compounding methods previously employed in the breast and significant increases in lesion and structural conspicuity are noted due to a reduction in speckle noise. The improvements from compounding in 3D, surface-rendered Doppler imaging of vasculature are striking.

THE 3D AND 2D COLOR FLOW DISPLAY OF BREAST MASSES

A prospective study was performed in 24 women with breast masses on mammography going on to surgical biopsy. 2D and 3D power mode and frequency shift color flow Doppler scanning and display were compared. Vessels were displayed as rotatable color volumes in 3D, superimposed on gray-scale slices. The latter were stepped sequentially through the imaged volume. Radiologists rated the masses in each display (3D, 2D and videotapes) on a scale of 1 to 5 (5 = most suspicious) for each of six conventional gray-scale and six new vascular criteria. Thirteen masses proved to be benign and 11 were malignant. 3D provided a stronger subjective appreciation of vascular morphology and allowed somewhat better ultrasound discrimination of malignant masses than did the 2D images or videotapes (specificities of 85%, 79% and 71%, respectively, at a sensitivity of 90%). Only in 3D did the vascularity measures display a trend towards significance in this small study.

3-D color Doppler image quantification of breast masses

In this article, new measures obtained from color Doppler images are introduced and a pilot study is described, in which these and previously published indices are evaluated for use in future work. Twenty women with breast masses observed on mammography and going to surgical biopsy were studied. Of the masses, 11 proved to be benign and 9 were malignant. Both 3-D mean frequency shift (f-CDI) and power mode Doppler (p-CDI) imaging were performed. To identify the mass and other regions of interest, vessels were displayed as rotatable 3-D color volumes, superimposed on selectable grey-scale/color flow slices. Doppler signals were recorded in each of 6 ellipsoidal regions of interest in and around the mass and 2 in normal tissues. Seven measures were computed in each region, three from power mode, two from mean frequency and two from combinations of both. Radiologists rated the grey-scale appearances of the masses on a scale of 1 to 5 (5=most suspicious) for each of 6 conventional grey-scale criteria. Of the individual vascularity measures in individual ROIs, the log speed-weighted pixel density and log power-weighted pixel density in the lesion internal periphery showed the greatest discrimination of malignancy, although neither was statistically significant nor as good as the peak variables described below. The mean visual grey-scale rating was the best discriminator overall, but two peak vascularity measures each made promising scatterplots in conjunction with the average visual grey-scale rating. These two vascularity measures were the log peak normalized power-weighted pixel density (peak NPD) and log of peak mean Doppler frequency times the peak NPD (vM x NPD(M)). Each of these two values was the maximum in any one of the five chosen ROIs closely associated with the mass. A possible rationale for the relative success of these peak values is the blood signals normalization and the inhomogeneity of most breast cancers and the expectation that the highest velocities (shunting) and largest collections of blood are not necessarily in the same region in and around the tumor. Peak NPD of cancers varied with age, decreasing by a factor of 45 from 33 to 77 y.

Analysis of three-dimensional Doppler ultrasonographic quantitative measures for the discrimination of prostate cancer.

The purpose of this study was to determine whether several quantitative ultrasonographic measures have potential to discriminate prostate cancer from normal prostate and to determine the best combination of these measures. The true spatial distributions of cancer within the prostates studied were obtained histologically after radical prostatectomy. The relationship between Doppler ultrasonography and microvessel count was also investigated.

Doppler quantitative measures by region to discriminate prostate cancer.

This study was conducted to assess if sonographic discrimination between healthy and cancerous prostate tissue might be improved using regional analysis of ultrasound (US) Doppler measures. A total of 39 subjects underwent 3-D Doppler sonography before radical prostatectomy. Cancer locations were identified from hematoxylin and eosin (H&E)-stained slides. Three prostate data volumes consisting of a frequency shift and power-mode Doppler US and whole mount histology images were spatially registered for each prostate, then divided into entirely 1 mL-sized regions of cancerous or noncancerous tissue. Each prostate was visually divided into a peripheral and a periurethral region within which US Doppler measures were calculated. Receiver operating characteristic (ROC) and simulated biopsy analyses within each prostate were performed. Mean speed in colored pixels (V), and speed-weighted pixel density (SWD) are good discriminators for prostate cancer in the periurethral and the peripheral regions, respectively. Using SWD in a simulated biopsy yields increased cancer detection in the peripheral region.

Approximate quantification of detected fractional blood volume and perfusion from 3-D color flow and Doppler power signal imaging

Ultrasound systems have been quite limited in the degree to which they facilitate quantification of reasonably available detected mean tumor flow characteristics such as: detected fractional blood volume, B v; speed-weighted, detected fractional blood volume, Bvv and mean speed of regionally-detected blood flow, v¯. This lack of quantification follows from the relative disinterest in such quantitative imaging as well as the dependence of any backscattered signal levels on intervening tissues and physical barriers to discrimination of the very slow flow and weak scattering amplitude of the dominant (capillary) blood pool relative to the soft tissue motion and scattering level, respectively. However, new signal power imaging modes offer some of the necessary data for measures such as Bv. Some color flow imaging systems also now offer improved clutter cancellers, sharper high pass filters, and other flexibility in adjustments for low speed flow. In addition, intravenous contrast agents will soon become available. Doing the best that is reasonably possible to quantify observed flow characteristics should aid objective assessment of the potential role of color flow and power mode ultrasound in utilizing the high metabolic rate of most breast cancers and metastases. Examples on a benign breast mass are presented

Analysis of three-dimensional ultrasound Doppler for the detection of prostate cancer☆

OBJECTIVES To investigate the relative effectiveness of Doppler ultrasound quantitative measures in discriminating prostate cancer from normal prostate tissue. The true locations of prostate cancer within these prostates were determined by histologic examination after radical prostatectomy. METHODS Three-dimensional Doppler ultrasound data were acquired from 39 men before radical prostatectomy. The removed prostates were sectioned and all cancerous regions in each prostate were identified on whole-mount hematoxylin-eosin-stained slides. The ultrasound and histologic data were then spatially registered. Biopsy results were simulated on a grid of potential sites within each prostate. Along each simulated biopsy site, the amount of cancer was computed from the hematoxylin-eosin-identified cancerous regions and the peak speed-weighted pixel density (SWD) was compared. RESULTS By selecting the biopsy sites with higher associated SWDs within each sextant, the probability of having at least one positive biopsy within a prostate increased from 75% if the SWD was ignored to 85% if only the top 15% of potential biopsy sites in each sextant were selected. This trend was seen within each sextant individually as well. CONCLUSIONS Doppler ultrasound provides discriminatory information for prostate cancer using the SWD. Translating this into a practical strategy that might improve the yield of prostate biopsy remains under development. The results of our study indicate that biopsying regions of high Doppler color could potentially increase the cancer yield to a small degree and improve the accuracy of the biopsy results. These results also objectively verify previous visual studies suggesting a modest improvement with the use of color Doppler.

The Use of Ultrasound Mean Acoustic Attenuation to Quantify Bone Formation during Distraction Osteogenesis Performed by the Ilizarov Method: Preliminary Results in Five Dogs

Daniel BL, Waanders NA, Zhang Y, Moskalik A, Fowlkes JB, Rubin JM, Goulet JA, Adler RS. The use of ultrasound mean acoustic attenuation to quantify bone formation during distraction osteogenesis performed by the Ilizarov method: preliminary results in five dogs. RATIONALE AND OBJECTIVES.Management of distraction during Ilizarov limb lengthening remains primarily clinical and empirical. Estimates of tissue acoustic attenuation were evaluated for their ability to quantify bone formation within the distraction gap. METHODS.Five dogs had tibias lengthened by the method of Ilizarov. Mean acoustic attenuation measurements at multiple positions across the distraction gap were compared with corresponding x–ray computed tomography attenuation measurements. RESULTS.Computed tomography and ultrasound attenuation displayed similar quantitative behavior across the gap. Linear correlation between them ranged from R2 = .878 to R2 = .131. Fibrous interzone width estimates based on computed tomography and ultrasound attenuation measurements were correlated, based on our preliminary data with R2 = .519. These estimates are independent of the width of distraction. CONCLUSIONS.Ultrasound parallels computed tomography as a measure of bone formation within the distraction gap. Future studies are needed to improve acoustic attenuation data acquisition techniques and to evaluate their potential as a tool for optimizing early distraction rates in patients at risk for raterelated complications.

3D registration of ultrasound with histology in the prostate

The authors hypothesize that frequency shift and power mode Doppler imaging in 3D can provide valuable diagnostic information for the detection and staging of prostate cancer. 3D data sets of both Doppler modes are taken of each prostate prior to radical prostatectomy. An in vitro b-mode scan is performed after surgery. The prostate is then cut into 3 mm sections and H&E and factor 8 stained whole mount slides are made. The in vitro scan and the sectioning are done in the same orientation, so that the slides can easily be aligned with the in vitro scan. The boundary of the prostate is fit with an ellipsoid and the urethra is identified in the two Doppler scans and the stacked histology slides, providing the scaling, translational, and rotational components of an affine transform. The resulting transform is used to register the volumetric data sets from a particular prostate, which allows quantitative vascular measures to be made in and around regions of histologically identified cancer. It can also be used to align data from different prostates to produce a probability map of where cancer occurs and a vascularity density map of cancerous prostates.

A Hand-Controlled, 3D Ultrasound Guide and Measurement System

Three dimensional (3D) ultrasound images provide superior visualization of anatomical features than two dimensional (2D) power mode or frequency shift color Doppler B-scan ultrasound images. However, 3D data sets require transducer position measurements that can be problematic. One previously investigated approach completely automates transducer positioning and measurements with one dimensional motorized stages. Other system types permit full six degree-of-freedom (Dof) manual transducer manipulation while measuring all six axes. In contrast this work explores the middle ground of measuring the two degrees of freedom most important for free-hand scanning while fixing the remaining axes for minimized complexity.