Using machine learning to disentangle heterogeneity within and between psychosis and depression: Improving pathways for precision medicine in psychiatry

Download Statistics

Downloads

Downloads per month over past year

Lalousis, Paris Alexandros ORCID: https://orcid.org/0000-0001-9642-3080 (2022). Using machine learning to disentangle heterogeneity within and between psychosis and depression: Improving pathways for precision medicine in psychiatry. University of Birmingham. Ph.D.

Preview

Lalousis2022PhD_Redacted.pdf
Text - Redacted Version
Available under License All rights reserved.
Download (6MB) | Preview

Abstract

The aim of this PhD was to examine the clinical and biological -primarily structural brain- heterogeneity within and between depression and psychosis and provide tools for the improvement of diagnosis and targeted treatment. In chapter 3, a systematic review of structural neuroimaging studies in depression and psychosis identified potential transdiagnostic patterns of gray matter volume (GMV) and white matter volume (WMV) reductions in areas including the middle frontal gyrus, hippocampus, and left-sided posterior subgenual prefrontal cortex. In chapter 4, clinical/neurocognitive and neuroanatomical support vector machine (SVM) learning models demonstrated separability of prototypic depression from psychosis. Psychosis patients with affective comorbidity aligned more strongly to depressive rather than psychotic disease processes. In chapter 5, we identified two transdiagnostic neuroanatomically informed clusters which are clinically and biologically distinct, challenging current diagnostic boundaries in recent onset mental health disorders. In chapter 6, five clusters of schizophrenia with distinct immune signatures, associated with differing GMV and neurocognitive function were identified, with potential to inform the development of novel, targeted treatments. Overall, machine learning was utilised to elucidate and reduce heterogeneity within and between psychosis and depression, and identify biologically relevant and transdiagnostic subtypes that could become potential candidates for targeted treatment. The results are promising and challenge the current nosological system.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):