Development of Deep Belief Networks to support the diagnosis of psychiatric disorders from large-scale public databases

• Grant number: 15/14993-7
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: November 01, 2015
• End date: April 30, 2016
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: João Ricardo Sato
• Grantee: Walter Hugo Lopez Pinaya
• Supervisor: Orla Doyle
• Host Institution: Centro de Matemática, Computação e Cognição (CMCC). Universidade Federal do ABC (UFABC). Ministério da Educação (Brasil). Santo André , SP, Brazil
• Institution abroad: King's College London, England
• Associated to the scholarship: 13/05168-7 - Development of Deep Belief Networks to support the diagnosis of psychiatric disorders, BP.DR

Abstract

Neuroimaging studies in psychiatry can potentially elucidate biological pathways, identify biomarkers of disease, prognosis, or treatment, and help redefine diagnostic boundaries. However, the availability of large-scale dataset is of paramount importance to properly evaluate these studies. In addition, the high-dimensional space of neuroimaging data requires sophisticated processing methods to analyze and establish meaningful conclusions. The Deep Belief Network is a machine learning method that exploit the unknown input data structure to extract multiple levels representations. This internship project will apply this multivariate approach to investigate the regularities of large-scale neuroimaging datasets. The intern will also evaluate Deep Belief Network based method as a predictive diagnosis model, comparing its performance with others studies from the literature. This project will explore several public datasets addressing different clinical conditions, as schizophrenia, bipolar disorder, autism spectrum, and ADHD. The internship will be at the King's College London under the supervision of Dr. Orla M. Doyle.