Grant: PJT PA: ICR - Lev. AI to trans. canc. rese. - AI to accel. Breast Can. Research

Lymphedema affects up to 25% of breast cancer and 17% of cancer survivors. Breast Cancer-Related Lymphedema (BCRL) is marked by chronic arm swelling that some survivors report as being worse than the cancer itself. BCRL is a leading cause of potentially life-threatening infections of the skin and subcutaneous tissues. Diagnosing BCRL can be difficult even for the most experienced clinicians due to its subtle signs, an issue amplified in remote regions. Early diagnosis, however, enables effective management through surgical or non-invasive interventions. The main goal of this project is to develop AI models to diagnose and stage BCRL from ultrasound data, without relying on expert radiologists. A variety of diagnostic tests and physical assessments at our Center of Excellence lymphedema clinic will be used to accurately stage BCRL, forming the training database for the AI model. Additionally, we will track patients over time, allowing the AI to predict how the condition may progress (something even top clinics struggle to do today). To increase accessibility and reduce costs, we will use battery-operated pocket-size ultrasound machines. We also aim to train AI models that mimic clinical reasoning, enabling clinicians to understand AI decisions and effectively communicate them to patients. A key outcome is the publication of a unique dataset, which will enable training AI models to identify tissue properties that contribute to BCRL. Breast centres and lymphedema services at our two partner hospitals will act as early adopters to enable prospective validation, workflow integration, and iterative refinement. Dr. Towers will liaise with the International Society of Lymphology to communicate results that may inform future updates to lymphedema staging criteria.