11.30 – 12.00 jakob kather - Predicting oncogenic driver mutations directly from histology with deep learning

Precision oncology is guided by molecular and genetic features of tumors. However, widespread implementation into clinical practice is limited because these biomarkers are costly, require significant expertise and are tested on tumor tissue, which is highly limited and expensive to handle. Virtually every cancer patient gets a tissue biopsy or tumor excision as part of the diagnostic workup and this tissue is routinely stained with hematoxylin and eosin. We have developed a deep learning-based technology to predict molecular features, prognosis and markers for treatment response directly from these ubiquitously available images. We and others have shown that deep learning can find patients with multiple actionable alterations in clinically relevant tumor types. In this talk, I will give an overview of the current performance of these methods and discuss future directions for deep learning in histology from a clinical point of view.

CV Jakob Nikolas Kather is a physician-scientist at RWTH Aachen University Hospital and the German Cancer Research Center, Heidelberg, Germany. He specializes in Gastrointestinal Oncology and heads an interdisciplinary team of computational and clinical researchers applying Machine Learning to routine clinical data, especially histological images. In multiple publications, he has bridged histopathology, cancer immunotherapy and machine learning with the aim of implementing computer-based decision support in clinical workflows.