The effect of breast density on the performance of deep learning-based breast cancer detection methods for mammography

Dustler M, Dahlblom V, Tingberg A, Zackrisson S
Proc. SPIE 11513, 15th International Workshop on Breast Imaging (IWBI2020), 1151324 (22 May 2020); doi: 10.1117/12.2564328

Aim and Objective
This study analyses whether there are correlations between BIRADS breast density classification and Transpara® Exam Scores, a deep learning-based artificial intelligence (AI) decision support system.

Materials and Method
A set of 13838 mammography screening exams were used. All cases had BIRADS density values available. The set included 2304 exams (11 cancers) in BIRADS 1, 5310 (51 cancers) in BIRADS 2, 4844 (73 cancers) in BIRADS 3 and 1223 (22 cancers) in BIRADS 4. Transpara (version 1.4.0), assigned a 1-10 risk score to each screening exam denoting the likelihood of cancer. Risk scores were separately recorded for the four density categories for cancer
cases and non-cancer cases respectively.

Results

The results suggest that Transpara risk categorization is not affected by density, as though some density categories receive higher risk assessments in general, this does not hold for cancer cases, which show uniformly high risk values despite density

Conclusion
This study shows the potential for Transpara to improve screening sensitivity, even for women with high breast density.