KEYWORDS: Cancer, Breast, Breast cancer, Mammography, Cancer detection, Deep learning, Medical physics, Visualization, Risk assessment, Digital mammography
When developing Deep Learning models intended for clinical applications, understanding which part of the input contributed the most to the final decision is crucial. Our study brings interpretability to a Breast Cancer Risk (BCR) prediction by exploring whether the model relies on the laterality of the breast, where cancer ultimately develops, and how this reliance evolves over time. A dataset of 1210 Full-Field-Digital-Mammography exams with 0 to 7 Years To Cancer was used. MIRAI model was employed for BCR predictions. To determine which side of the breast contributed the most to the BCR prediction, the signal difference between left and right breasts was calculated for eight attribution-based interpretability techniques. AUC was calculated to investigate whether the BCR prediction is predominantly made from the breast, where the cancer ultimately develops. For 0 to 1 Years To Cancer, the model predominantly predicts BCR based on the side of the breast where the cancer is already present AUC=0.92 to 0.95. The top-performing attribution methods achieved an AUC of 0.70 for mammograms captured 1 to 3 Years To Cancer. For exams that were 3 to 5 Years To Cancer, a significant drop to AUC of 0.57 was observed. When moving to 5 to 7 Years To Cancer, focus on the breast with future cancer becomes random. All attribution methods showed that BCR predictions extending beyond three years from screen-detected cancer are most likely based on typical breast characteristics, such as density and other long-standing tissue patterns; however, for short-term BCR predictions, the model seems to detect early signs of tumor development.
Aim: To develop and subsequently perform a systematic study on the impact of parameter settings on the biological reproducibility and sensitivity of extracted radiomic features from Full Field Digital Mammography (FFDM) images for the task of Breast Cancer Risk assessment. Methods: Cranio-caudal (CC) ”FOR PRESENTATION” images (88 in total, two centers: Slovenia and Belgium) were used for this study. Biological reproducibility of radiomic features was evaluated with two tests: reproducibility of extracted features between left and right breasts and by reproducibility of extracted features between the original and 4 perturbed images. The quantification was done using the intra-class correlation (ICC) coefficient between values of extracted radiomic features. To determine biological sensitivity, AUC between groups with low and high breast cancer risk was calculated. For the selection of optimal radiomic feature parameters, thresholds of 0.75 and 0.7 were defined for ICC and AUC, respectively. Results: Parameters bin Count and distances highly influenced biological reproducibility and sensitivity of specific radiomic features. Parameters weightingNorm and symmetricalGLCM had no effect. Overall, only 12/93 radiomic features passed the reproducibility and sensitivity tests in both centers. For five of these features, parameter ranges were crucial. Reproducibility varied greatly between the centers of Belgium and Slovenia. Conclusions: Rather than single radiomic parameters, parameter ranges were found to be a reasonable description for acceptable biological reproducibility and sensitivity. Overall, 12/93 radiomic features were found to be potential candidates for breast cancer risk prediction tasks, however further analysis is needed before definitive recommendations can be made.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.