Abstract
The current diagnostic practice in psychiatry lacks objective markers for predicting psychiatric diagnosis and illness course to improve the consistency in the diagnostic process and facilitate more precise prevention and treatment of mental disorders.1 In addition, studies have shown that a significant association between a disease and a feature does not necessarily mean it can be used to predict the disease clinically.2,3 Therefore, with the increasing amounts of data available, new opportunities have emerged to apply cutting-edge methods such as deep learning (DL) to build prediction models using high-dimensional data to capture nonlinear relationships spanning multiple data sources (multimodal).1,4,5 Currently, machine learning (ML), and specifically the DL subspeciality, have been used to establish clinically applicable prediction tools within other medical fields, such as breast cancer diagnostics, decision-making during the COVID-19 pandemic, and prediction of cardiac arrest from emergency calls.6-13

This publication has 41 references indexed in Scilit: