Since machine learning was brought to psychiatric research, the use of classification models for clinical diagnoses has received a lot of attention, although there are still some practical difficulties. The use of machine learning to recognise neurobiological traits and offer fresh perspectives on the nosology of mental diseases has become popular in the meantime.
A research team at Zhejiang University’s Department of Psychology and Behavioral Sciences under the direction of Assistant Professor Chen Ji summarised several new directions based on earlier research on psychiatric machine learning and highlighted the critical function of machine learning in supplying neurobiological and nosological insights into mental disorders. In the journal Biological Psychiatry, a review article presenting their findings was published.
Chen Ji et al. offered their opinions by examining the most recent developments in the field of machine learning for mental illnesses.
(1) The biological characteristics closely associated to the pathophysiological processes of mental diseases can be identified by using the classification accuracy rate of machine learning models as a dependent variable;
(2) It is possible to examine the DSM’s taxonomic relationships among mental diseases using the categorization models’ accuracy rates;
(3) It is possible to examine the dimensional (within diagnosis) and frequently overlapping (across diagnoses) symptomatology of psychiatric illness using semi-supervised and unsupervised machine learning.
The researchers proposed frequent mistakes related to input data or analytic procedures at the technical level, concentrating on the idea of “garbage-in, garbage-out,” in order to address these methodological viewpoints and application tactics.
This study offers a fresh viewpoint for identifying mental disease biomarkers and for addressing the problems of diagnostic heterogeneity and co-morbidity in psychiatric research. Additionally, it emphasises the significance of resolving probable issues and hazards in data collection, model development, and result elaboration before employing machine learning to find trustworthy biomarkers and investigate novel taxonomies for mental diseases.
