The overall objective of my projects has been to accurately predict recurrence risk in breast cancer patients from pathology slides using computational imaging and machine learning. Knowledge of a patient’s risk of recurrence is important, as it helps determine what treatment they should or should not receive. For instance, a patient may have a prognosis so favorable that the benefit of chemotherapy is likely to be very low. In such a case, the patient should probably not receive chemotherapy.
In this talk, I will discuss: computational imaging techniques, such as Fourier ptychography, used for enhanced image acquisition; machine-learning algorithms applied to breast cancer pathology images in order to predict patient recurrence risk; and an end-to-end system under development to aid pathologists in clinical decision-making.