Computer Aided Diagnosis for Breast Ultrasound Imagery (CADBUSI)
A joint project between Mayo-La Crosse and University of Wisconsin-La Crosse. Supported by a grant from Mayo Health.
Goal
Produce software that performs at least as well as human experts in the accurate interpretation and clinical assessment of breast lesions detected via ultrasound. The image below is a rough roadmap of our design.
More details can be found in the project proposal and these presentation slides.
Our focus is on building an unintrusive system that provides:
- decision support focused on BIRADS 3 (no biopsy) and BIRADS 4 (biopsy required) lesions
- an overall score / probability of malignancy / BIRADS score
- explainable results
- BIRADS characteristics
- heatmap overlay showing classification relevant parts of image
- an interactive mechanism for the radiologist to refine the diagnosis