Machine Learning in Digital Pathology
Abstract
Digital pathology produces exactly the kind of high-volume, high-dimensional data that machine learning is good at, but clinical adoption depends on more than predictive accuracy. This talk covers our work on machine learning for medicine and diagnostics, the interpretability and fairness requirements that clinical deployment imposes, and the step from a predictive model to a decision support system a clinician can actually act on.
I gave an invited talk at the Roche Digital Pathology Summit on the use of machine learning in medical and diagnostic settings.
The talk drew on our machine learning for medicine research — hospital readmission prediction, biomarker analysis, and diagnostic modelling — and focused on what has to be true of a model before it can be deployed clinically: interpretability, fairness, and the transition from a prediction to a decision a clinician can act on.
Sandro Radovanović
Assistant Professor at University of Belgrade
My research interests include machine learning, development and design of decision support systems, decision theory, and fairness and justice concepts in algorithmic decision making.