Abstract:

Depression, a complex and heterogeneous mental disorder, leads to great global burdens of disease. Although diagnosis based on nosology is broadly used in several domains, it is still unable to direct the exploration of pathological mechanism of depression. In addition, several treatments developed by this diagnosis have poor outcomes due to its low prediction validity. Computational approaches to psychiatry remedy those limitations and help to improve understanding, prediction and treatment for depression by two complementary approaches: data-driven and theory-driven. Theory-driven approaches apply models to multiple levels of analysis from the prior knowledge or hypojournal of depression. Data-driven approaches, however, adopt machine-learning methods to analyze high-dimensional data to improve the diagnostic and predictive accuracies of depression, and eventually, promote the treatment effects. With the development and combination of these two approaches as well as the integration of resources, it is promising to cure depression and prevent it from occurrence.

Key words: depression, computational psychiatry, computational models, machine learning, diagnosis, treatment