Metastatic breast cancer detection
Signalling patients who develop metachronous metastases after primary breast cancer.
Due to the vast amount of survivors, it is infeasible for a cancer registry to manually follow-up and register new signs of progression for all patients. Hence, we trained a machine learning model based on health declaration data to detect which individuals may have developed metastatic cancer. This resulted in a more complete cancer registry and thereby offered better support for oncological research.
Results:
Created a probabilistic match between cancer registry data and a health declarations database.
Trained a machine learning classifier to predict the development of metastatic cancer with an average precision of 0.95.