Machine learning for prediction and visualisation of brain diseases. Demonstration on Alzheimer's disease

Abstract

Razvan Marinescu is a postdoctoral researcher at MIT, working with Polina Golland to develop machine learning algorithms for quantifying pathologies in the human brain, as well as visualisation of machine learning prediction models. During his PhD, he developed Bayesian statistical models for prediction of Alzheimer’s disease and related neurodegenerative diseases. He further organised TADPOLE, an international competition that gathered 92 algorithms from 33 teams for predicting the evolution of Alzheimer’s disease. At MIT, he currently works on methods to visualise what black-box classifiers learned, which are useful for debugging, increasing model performance and building trust in order to deploy such systems to the clinic.