A review on machine learning and deep learning approaches for depressive disorder diagnosis based on EEG and ERP signals

Abstract

Depression has become one of the most common disorders worldwide, affecting both individuals and societies, and objective diagnostic approaches are required to help with this challenge. Machine learning (ML) and deep learning (DL) techniques have become increasingly applied to find reliable markers of depressive symptoms based on electroencephalography (EEG) and event-related potentials (ERPs) data. In this review, we deliver a comprehensive analysis of the 69 published articles from 2020 to 2026 that exhibit ML/DL methods for recognizing depression based on EEG and ERP datasets. The studies reported include analysis at levels of the analytics pipeline such as signal pre-processing, features extraction, features selection and classification. Feature extraction is one of the main categories, which is further classified into three subcategories: Handcrafted features, Hybrid feature extraction and End to End feature extraction through deep learning. This review also evaluates the performance of the models in terms of evaluation metrics, key applications and achievements, introduces widely used publicly available datasets, examines clinical, ethical aspects, and patient perspectives, and finally provides guidance for researchers in this field by pointing out challenges, research gaps, and suggestions for future directions.