Mark Sak

Breast density measurements with ultrasound tomography: A comparison with film and digital mammography

PURPOSE To investigate the use of the whole-breast sound speed measurement as a marker of breast density (BD), a known risk factor for breast cancer. METHODS As part of an ongoing study of breast cancer detection, 249 patients were scanned with a clinical prototype that operates on the principles of ultrasound tomography. Typically, 40-100 sound speed tomograms were reconstructed from the scan data, corresponding to the entire volume of the breast of each patient. The data were used to estimate the volume averaged sound speed (VASS) of the breast for each patient. The corresponding mammograms were used to calculate mammographic percent density (MPD) using CUMULUS software. Film mammograms were available for 164 patients while 85 digital mammograms were available for the remaining patients. Standard statistical techniques were used to determine associations of breast sound speed with a variety of mammographic measures such as percent density, area of dense tissue, and area of nondense tissue. Furthermore, associations of breast sound speed with continuous variables such as age and weight and dichotomous variables such as parity and menopausal status were also assessed. RESULTS VASS was found to be significantly associated with MPD. The Spearman correlation coefficient (r(s)) between VASS and MPD was found to be 0.77 and 0.71 for film and digital mammography, respectively. VASS was positively correlated with dense areas by mammography, both digital (r(s) = 0.46) and film (r(s) = 0.56). VASS was negatively associated with nondense area by mammography, both digital (r(s) = -0.58) and film (r(s) = -0.63). BD by all methods was less in postmenopausal than in premenopausal women. The MPD was lower in the postmenopausal group (by 6.6%, p < 0.08, for the digital group and 7.73%, p < 0.007, for the film group). The VASS was also lower in the postmenopausal group (by 15 m∕s, p < 0.001 for the digital group and 8 m∕s, p < 0.08, for the film group). The association of MPD with age was characterized with r(s) = -0.06 (p < 0.6) for digital mammography and r(s) = -0.53 (p < 0.002) for film mammography. For weight, the MPD associations were characterized by r(s) = -0.53 (p < 0.0001) for digital mammography and -0.38 (p < 0.0001) for film mammography. The association of VASS with age was r(s) = -0.33 (p < 0.002) for the digital group and -0.17 (p < 0.03) for the film group. For weight, the relationship was characterized with r(s) = -0.45 (p < 0.001) for the digital group and -0.37 (p < 0.0001) for the film group. CONCLUSIONS The association between VASS and MPD is strong for both film and digital mammography, suggesting that VASS is a viable measure of breast density. This result sets the stage for future work that will focus on directly testing the association of VASS with breast cancer risk.

Comparison of 3D and 2D breast density estimation from synthetic ultrasound tomography images and digital mammograms of anthropomorphic software breast phantoms

Breast density descriptors were estimated from ultrasound tomography (UST) and digital mammogram (DM) images of 46 anthropomorphic software breast phantoms. Each phantom simulated a 450 ml or 700 ml breast with volumetric percent density (PD) values between 10% and 50%. The UST based volumetric breast density (VBD) estimates were calculated by thresholding the reconstructed UST images. Percent density (PD) values from DM images were estimated interactively by a clinical breast radiologist using Cumulus software. Such obtained UST VBD and Cumulus PD estimates were compared with the ground truth VBD values available from phantoms. The UST VBD values showed a high correlation with the ground truth, as evidenced by the Pearson correlation coefficient of r=0.93. The Cumulus PD values also showed a high correlation with the ground truth (r=0.84), as well as with the UST VBD values (r=0.78). The consistency in measuring the UST VBD and Cumulus PD values was analyzed using the standard error of the estimation by linear regression (σE). The σE value for Cumulus PD was 1.5 times higher compared to the UST VBD (6.54 vs. 4.21). The σE calculated from two repeated Cumulus estimation sessions (σE=4.66) was comparable with the UST. Potential sources of the observed errors in density measurement are the use of global thresholding and (for Cumulus) the human observer variability. This preliminary study of simulated phantom UST images showed promise for non-invasive estimation of breast density.

Current and future methods for measuring breast density: a brief comparative review

Breast density is one of the strongest predictors of breast cancer risk. Women with the densest breasts are 4 to 6 times more likely to develop cancer compared with those with the lowest densities. Breast density is generally assessed using mammographic imaging; however, this approach has limitations. Magnetic resonance imaging and ultrasound tomography are some alternative imaging modalities that can aid mammography in patient screening and the measurement of breast density. As breast density becomes more commonly discussed, knowledge of the advantages and limitations of breast density as a marker of risk will become more critical. This review article discusses the relationship between breast density and breast cancer risk, lists the benefits and drawbacks of using multiple different imaging modalities to measure density and briefly discusses how breast density will be applied to aid in breast cancer prevention and treatment.

Relationship between breast sound speed and mammographic percent density

Despite some shortcomings, mammography is currently the standard of care for breast cancer screening and diagnosis. However, breast ultrasound tomography is a rapidly developing imaging modality that has the potential to overcome the drawbacks of mammography. It is known that women with high breast densities have a greater risk of developing breast cancer. Measuring breast density is accomplished through the use of mammographic percent density, defined as the ratio of fibroglandular to total breast area. Using an ultrasound tomography (UST) prototype, we created sound speed images of the patients breast, motivated by the fact that sound speed in a tissue is proportional to the density of the tissue. The purpose of this work is to compare the acoustic performance of the UST system with the measurement of mammographic percent density. A cohort of 251 patients was studied using both imaging modalities and the results suggest that the volume averaged breast sound speed is significantly related to mammographic percent density. The Spearman correlation coefficient was found to be 0.73 for the 175 film mammograms and 0.69 for the 76 digital mammograms obtained. Since sound speed measurements do not require ionizing radiation or physical compression, they have the potential to form the basis of a safe, more accurate surrogate marker of breast density.

Breast tissue composition and breast density measurements from ultrasound tomography

It is known that breast cancer risk is greater in women with higher breast densities. Currently, breast density is measured using mammographic percent density, defined as the ratio of fibroglandular to total breast area on a two dimensional mammogram. Alternatively, systems that use ultrasound tomography (UST) create tomographic sound speed images of the patients breast. These volumetric images can be useful as a diagnostic aid because it is also known that sound speed of tissue is proportional to the density of the tissue. The purpose of this work is to expand on the comparisons of the two imaging modalities by introducing new ultrasound tomography measurements that separate and quantify the fatty and dense tissue distributions within the breast. A total of 249 patients were imaged using both imaging modalities. By using k-means clustering, correlations beyond the volume averaged sound speed of the ultrasound images and the mammographic percent density were investigated. Both the ultrasound and mammographic images were separated into dense and fatty regions. Various associations between the global breast properties as well as separate tissue components were found.

2D/3D image fusion of x-ray mammograms with speed of sound images: evaluation and visualization

Breast cancer is the most common cancer among women. The established screening method to detect breast cancer is X-ray mammography. However, X-ray frequently provides poor contrast of tumors located within glandular tissue. In this case, additional modalities like MRI are used for diagnosis in clinical routine. A new imaging approach is Ultrasound Computer Tomography, generating three-dimensional speed of sound images. High speed of sound values are expected to be an indicator of cancerous structures. Therefore, the combination of speed of sound images and X-ray mammograms may benefit early breast cancer diagnosis. In previous work, we proposed a method based on Finite Elements to automatically register speed of sound images with the according mammograms. The FEM simulation overcomes the challenge that X-ray mammograms show two-dimensional projections of a deformed breast whereas speed of sound images render a three-dimensional undeformed breast in prone position. In this work, 15 datasets from a clinical study were used for further evaluation of the registration quality. The quality of the registration was measured by the displacement of the center of a lesion marked in both modalities. We found a mean displacement of 7.1 mm. For visualization, an overlay technique was developed, which displays speed of sound information directly on the mammogram. Hence, the methodology provides a good basis for multimodal diagnosis using mammograms and speed of sound images. It proposes a guidance tool for radiologists who may benefit from the combined information.

Efficient Segmentation of a Breast in B-Mode Ultrasound Tomography Using Three-Dimensional GrabCut (GC3D)

As an emerging modality for whole breast imaging, ultrasound tomography (UST), has been adopted for diagnostic purposes. Efficient segmentation of an entire breast in UST images plays an important role in quantitative tissue analysis and cancer diagnosis, while major existing methods suffer from considerable time consumption and intensive user interaction. This paper explores three-dimensional GrabCut (GC3D) for breast isolation in thirty reflection (B-mode) UST volumetric images. The algorithm can be conveniently initialized by localizing points to form a polygon, which covers the potential breast region. Moreover, two other variations of GrabCut and an active contour method were compared. Algorithm performance was evaluated from volume overlap ratios (TO, target overlap; MO, mean overlap; FP, false positive; FN, false negative) and time consumption. Experimental results indicate that GC3D considerably reduced the work load and achieved good performance (TO = 0.84; MO = 0.91; FP = 0.006; FN = 0.16) within an average of 1.2 min per volume. Furthermore, GC3D is not only user friendly, but also robust to various inputs, suggesting its great potential to facilitate clinical applications during whole-breast UST imaging. In the near future, the implemented GC3D can be easily automated to tackle B-mode UST volumetric images acquired from the updated imaging system.

Comparison of sound speed measurements on two different ultrasound tomography devices

Ultrasound tomography (UST) employs sound waves to produce three-dimensional images of breast tissue and precisely measures the attenuation of sound speed secondary to breast tissue composition. High breast density is a strong breast cancer risk factor and sound speed is directly proportional to breast density. UST provides a quantitative measure of breast density based on three-dimensional imaging without compression, thereby overcoming the shortcomings of many other imaging modalities. The quantitative nature of the UST breast density measures are tied to an external standard, so sound speed measurement in breast tissue should be independent of specific hardware. The work presented here compares breast sound speed measurement obtained with two different UST devices. The Computerized Ultrasound Risk Evaluation (CURE) system located at the Karmanos Cancer Institute in Detroit, Michigan was recently replaced with the SoftVue ultrasound tomographic device. Ongoing clinical trials have used images generated from both sets of hardware, so maintaining consistency in sound speed measurements is important. During an overlap period when both systems were in the same exam room, a total of 12 patients had one or both of their breasts imaged on both systems on the same day. There were 22 sound speed scans analyzed from each system and the average breast sound speeds were compared. Images were either reconstructed using saved raw data (for both CURE and SoftVue) or were created during the image acquisition (saved in DICOM format for SoftVue scans only). The sound speed measurements from each system were strongly and positively correlated with each other. The average difference in sound speed between the two sets of data was on the order of 1-2 m/s and this result was not statistically significant. The only sets of images that showed a statistical difference were the DICOM images created during the SoftVue scan compared to the SoftVue images reconstructed from the raw data. However, the discrepancy between the sound speed values could be easily handled by uniformly increasing the DICOM sound speed by approximately 0.5 m/s. These results suggest that there is no fundamental difference in sound speed measurement for the two systems and support combining data generated with these instruments in future studies.

Automatic Segmentation of Ultrasound Tomography Image

Ultrasound tomography (UST) image segmentation is fundamental in breast density estimation, medicine response analysis, and anatomical change quantification. Existing methods are time consuming and require massive manual interaction. To address these issues, an automatic algorithm based on GrabCut (AUGC) is proposed in this paper. The presented method designs automated GrabCut initialization for incomplete labeling and is sped up with multicore parallel programming. To verify performance, AUGC is applied to segment thirty-two in vivo UST volumetric images. The performance of AUGC is validated with breast overlapping metrics (Dice coefficient (D), Jaccard (J), and False positive (FP)) and time cost (TC). Furthermore, AUGC is compared to other methods, including Confidence Connected Region Growing (CCRG), watershed, and Active Contour based Curve Delineation (ACCD). Experimental results indicate that AUGC achieves the highest accuracy (D = 0.9275 and J = 0.8660 and FP = 0.0077) and takes on average about 4 seconds to process a volumetric image. It was said that AUGC benefits large-scale studies by using UST images for breast cancer screening and pathological quantification.

Ultrasound tomography for breast cancer screening (Conference Presentation)

Both mammography and standard ultrasound (US) rely upon subjective criteria within the breast imaging reporting and data system (BI-RADS) to provide more uniform interpretation outcomes, as well as differentiation and risk stratification of associated abnormalities. In addition, the technical performance and professional interpretation of both tests suffer from machine and operator dependence. Breast MR has become the new gold standard for screening of high-risk women but has cost and access limitations in extending screening to the entire population. We have been developing a new technique for breast imaging that is based on ultrasound tomography which quantifies tissue characteristics while also producing 3-D images of breast anatomy. Results are presented from clinical studies that utilize this method. Informed consent was obtained from all patients, prospectively recruited in an IRB-approved protocol following HIPAA guidelines. Images were produced by tomographic algorithms for reflection, sound speed and attenuation. All images were reviewed by a board-certified radiologist who has more than 20 years of experience in breast imaging and US-technology development. In the first phase of the study, UST images were compared to multi-modal imaging to determine the appearance of lesions and breast parenchyma. In the second phase of the study, correlative comparisons with MR breast imaging were used to establish basic operational capabilities of the UST system including the identification and characterization of parenchymal patterns. Our study demonstrated a high degree of correlation of breast tissue structures relative to fat subtracted contrast enhanced MRI. With a scan duration of ~ 1-3 minutes, no significant motion artifacts were observed.