Abstract: We work on the breast imaging malignancy segmentation task while focusing on the train- ing process instead of network complexity. We designed a training process based on a modified U-Net, increasing the overall segmentation performances by using both, benign and malignant data for training. Our approach makes use of only a small amount of anno- tated data and relies on transfer learning from a self-supervised reconstruction task, and favors explainability.
Paper Type: methodological development
TL;DR: Two-step self- and fully-supervised training process for more precise malignancy segmentation in mammography
Track: short paper
Keywords: Mammography, Segmentation, Malignancy Detection, Explainability
Presentation Upload: zip
Presentation Upload Agreement: I agree that my presentation material (videos and slides) will be made public.