Using Deep Learning And Digital Pathology To Intrinsically Subtype Breast Cancer

Digital pathology is the reviewing of tissue slides on a computer monitor rather than using microscopes. It is gaining momentum with anatomical pathology laboratories transitioning to this technology. In addition to increasing efficiency of pathologists to generate reports for clinicians treating breast cancer, digital pathology allows the application of computer algorithms that objectively quantify and standardise results. Studies have found that algorithms applied to digitally scanned slides can provide molecular data from breast cancer cases. This would otherwise require molecular testing, which not all patients can access due to cost. We wish to develop advanced machine-learning algorithms to automatically identify digital signatures of genomic changes in invasive breast cancers from tissue slides. This will be used to improve equity of access to testing and improve prediction of prognosis and response to treatment. The ultimate goal is to allow more tailored patient therapies to give the best clinical outcome possible.

HOST INVESTIGATOR: Canterbury District Health Board