Using machine-learning approaches in non-invasive brain stimulation trials to identify clusters of symptoms associated with clinical response

• Grant number: 21/05237-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): July 01, 2021
• Effective date (End): June 30, 2022
• Field of knowledge: Health Sciences - Medicine - Psychiatry
• Principal Investigator: Andre Russowsky Brunoni
• Grantee: Dárin Oliveira Moura
• Host Institution: Hospital Universitário (HU). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated research grant: 19/06009-6 - Non-implantable neuromodulation therapies: a perspective for the depressed brain, AP.TEM

Abstract

The use of machine learning (ML) approaches has soared in Medicine, including in the Psychiatry and Psychology fields. Although the definition is broad, ML, a subfield of Artificial Intelligence (AI), usually involves using data-driven approaches in large datasets to enhance response prediction and diagnostic accuracy by iteratively adjusting parameters and features to achieve an optimal solution rather than delivering a fixed solution a priori. As open-access initiatives open space to merging and analyzing large datasets, and subjects in clinical trials are more and more assessed via multimodal approaches, understanding methods of handling Big Data is increasingly necessary to medical and psychology scientists regardless of their subfield. This scholarship is focused on presenting ML approaches to an undergrad student. The student will work together with his supervisor and a Postdoctoral fellow to acquire knowledge on different ML approaches and statistical packages. The student will develop his/her own research project, using the ELECT-TDCS dataset, a clinical trial that evaluated transcranial direct current stimulation (tDCS) efficacy in depression. The student will investigate whether specific symptoms, or cluster of symptoms, are associated with tDCS response. A similar approach than employed by Chekroud et al. to identify cluster of symptoms associated with antidepressant response will be used.