MACHINE LEARNING IN NEUROIMAGING: APPLICATIONS TO BRAIN AGING, AD, SCHIZOPHRENIA, AND BRAIN CANCER

Περίληψη

Quantitative and computational methods have increasingly provided insights in many neuroscience
problems. Amongst them, AI and machine learning have relatively recently emerged as very promising avenues
for knowledge discovery, especially in the era of complex, big and diverse data. Herein, applications
of machine learning in neuroimaging are discussed, with emphasis on aging and Alzheimer’s Disease (AD),
schizophrenia, and the most aggressive brain cancer, namely glioblastoma (GBM). In particular, machine
learning is shown to produce highly sensitive and specific imaging signatures of brain change during early
preclinical stages of AD, as well as to identify neuroanatomically distinct subtypes of schizophrenia. Finally,
machine learning is shown to produce imaging signatures that predict patient outcome. These representative
results highlight the potential of machine learning in neuroimaging as means to derive sensitive and
specific biomarkers, and to reduce complex and diverse data into a small number of dimensions capturing
different aspects of the neurobiology of brain diseases.