Reducing breast cancer overtreatment
Breast cancer is one of the leading causes of cancer death among women worldwide. Manifestation of breast cancer involves progression through different clinical and pathological phases that may finally culminate into metastatic disease. Even though it is often curable when caught early, there is a possibility of cancer recurrence many years after the first diagnosis. An accurate risk assessment of breast cancers or their precursors is paramount to deciding on the appropriate treatment regime.
In the AI for Oncology Lab, we are working to improve breast cancer risk assessment. In particular, several studies show that it is possible to avoid neoadjuvant therapy in more cases of early-stage breast cancers than what the current recurrence risk assessment models suggest. We are working to improve these risk assessment models by combining traditional clinical, pathological parameters and histopathological data using AI. Concurrently, we are working on an accurate risk assessment of ductal carcinoma in situ (DCIS), a potential precursor of breast cancer. Current risk models do not provide enough information to predict whether a DCIS lesion will progress into invasive breast cancer. Improving these models will lead to less overtreatment of breast cancer patients and a better quality of life.
In these projects we collaborate with the groups of Hugo Horlings, Ritse Mann and Jelle Wesseling.
- A. Panteli, J. Teuwen, H. Horlings, E. Gavves, "Sparse-shot Learning with Exclusive Cross-Entropy for Extremely Many Localisations", 2021-04-21
- S. Doyle, F. Dal Canton, J. Wesseling, C. I. Sánchez, J. Teuwen, "Mammary duct detection using self-supervised encoders", Medical Imaging 2022: Computer-Aided Diagnosis, 2022-04-04