Diagnosis of neurological and psychiatric diseases based on whole-brain functional connectivity using Machine Learning techniques

Abstract

The global problem this thesis aims to target is the inability of psychotherapists and psychiatrists always correctly to identify a presence of a mental illness. To give a constructive medical conclusion on a patient’s state, usually, it is not enough to only rely on symptoms concluded from a therapeutic session. Moreover, the diagnosis of that kind could be biased from both a therapist and a patient’s side. The former depends on the doctor’s knowledge and experience, and the latter is based on an ability to communicate the mental state. Notably, the more researchers investigate the cause of psychiatric diseases, the more they make sure that mental illnesses are developed due to specific changes in one’s brain. It could be the brain’s structure, functionality, or damage, leading to changes in a person’s behavior, thought process, interaction with other people, and sometimes difficulties in functioning as a healthy human being. It is believed that severe mental illnesses and neurological and developmental diseases result from abnormal connectivity in a brain network. That is why whole-brain functional connectivity is a significant source of information in this study. In simple words, it represents if and how the brain regions communicate with each other. This study presents generalized, usable, and reliable classification models to identify a specific neurological or psychiatric disease, Autism Spectrum Disorder and Schizophrenia, with 92.4% and 93.8% of accuracy respectfully, for further clinical application of the developed tool.