At this point, we can all be certain breast density is a real thing. But questions still remain about what constitutes dense breast tissue and how it should be measured.

Jim Pike is co-founder, president and CTO of VuCOMP, a company that develops automated breast density assessment technology and advanced computer-aided detection (CAD) for mammography. He spoke with DOTmed News about some of the benefits and challenges of a technology like this.

DMN: At one point the whole breast density discussion was quite contentious, but that seems to have softened. Why do you think this is the case, and is there much debate among radiologists about the value of breast density measurement at this point?

JP: Many doctors remain apprehensive about reporting breast density; patient groups continue to strongly advocate for it. Since clinical standards remain noncommittal about what should be done for women with dense breasts, reporting the information can cause difficulties for both the doctor and the patient.

In my opinion, doctors are recognizing the patient-driven desire for reporting breast density, and they seem to be slowly accepting its inevitability.

DMN: Describe what breast density actually is and what risks it presents to a woman.

JP: Breast density is the amount of dense (i.e. radio-opaque) fibroglandular tissue (as opposed to radiolucent fat tissue) in the breast. As breast density increases, the risk of developing breast cancer increases, while the sensitivity of mammography decreases. More specifically, mammographic sensitivity decreases by as much as 16 percent for dense breasts, and 40 percent of women will have dense breasts at some point in their lives. Additionally, women with a BI-RADS 4 breast density have over four times the relative risk of developing breast cancer compared to those with BI-RADS 1. Consequently, current guidelines (e.g. NCCN) suggest that additional screening modalities such as ultrasound or magnetic resonance imaging should be considered for some women with dense breasts.

DMN: How would a system like yours, an automated breast density assessment tool and CAD, inform the radiologist?

JP: M-Vu assesses breast density by examining the mammographic texture and appearance of the dense tissue, emulating -- quantitatively -- the approach advocated by the American College of Radiology. The advantage of this approach is that in addition to considering the percent of breast density, it also analyzes the texture and dispersion pattern of the tissue. Interestingly, in our own study, we noticed that in 4 percent of all pairs of cases in which the BI-RADS assignments (of the two cases) differed by one category, the difference in breast density (as assessed by radiologists) was inversely proportional to the difference in BI-RADS category. This suggests that percent breast density alone is insufficient to correctly assign BI-RADS categories to mammograms. Therefore, we designed our system to analyze the texture and dispersion pattern of tissue.

DMN: Do all breast density measurement systems use the same methodology?

JP: No, one approach, which is most often used, measures the physical absorption of X-rays (given by the pixel values in the mammogram) to estimate the percent volume covered by fibroglandular tissue. The percent volume is then mapped to a score from a to d, analogous to the BI-RADS category. Such volumetric methods have inherent limitations. First, pixel values in the mammographic images may not reflect the actual absorption of X-rays, decreasing the accuracy of -- or completely invalidating -- the volumetric assessments. For example, volumetric measurements from mammograms imaged using a Lorad Selenia are less accurate than those from the GE Senographe; the Selenia compression paddle can tilt during compression, whereas the Senographe paddle remains rigid. Furthermore, the pixel values of both standard for-presentation images and synthetic C-view images are not directly related to the X-ray attenuation. Second, since BI-RADS categories are defined not only by the percentage of the breast covered by dense tissue, but also by the dispersion of this tissue throughout the breast, volumetric methods are inherently limited in their ability to accurately reproduce BI-RADS scores.

DMN: You mentioned the BI-RADS categories — to what extent does the appearance-based approach agree with the category descriptors?

JP: For automated breast density systems to be efficacious in a clinical setting, they must yield accurate and consistent results. Initial studies support such efficacy. Gweon, et al.11, reported a weighted kappa statistic of 0.54 -- indicating moderate agreement -- between the Volpara density grade (Matakina Technology) and a consensus of three radiologists' BI-RADS categories. A similar analysis of M-Vu Breast Density (reported in an FDA clinical study) yielded a weighted kappa statistic of 0.64 -- indicating substantial agreement between the VuCOMP density grade and a consensus of 13 expert radiologists' BI-RADS categories. In a different study of 490 patients, the density estimates produced by M-Vu had less than half the variability of those produced by 10 radiologists.

DMN: What are you seeing out there in terms of interest and adoption?

JP: We have seen significant adoption from practices and clinics large and small who want to remove the subjectivity out of their breast density assessment. Sites are searching for an automated tool that provides consistent results from physician to physician and year over year. Radiologists are saying that VuCOMP's approach to analyzing structure, texture, and dispersion of fibroglandular tissue provides useful adjunctive information. There are now 17 states which have passed breast density notification laws and federal legislation is being developed now, so this has and certainly will continue to drive adoption.