Analysis of postpartum and other depression detection using machine learning and deep learning techniques: a...

Postpartum depression is a major pregnancy-related mental health issue. Currently, computational intelligence research is increasing rapidly for early detection. This systematic review provides a detailed analysis of previous studies based on datasets, models used in machine learning and deep learning, feature selection methods, preprocessing techniques, and performance metrics. Scholarly works were identified across various sources, and they focused on NLP approaches, multi-source data, and forecasting models. Findings show that the research area remains underdeveloped, with most investigations relying on small or single-site data and a small set of features. Although some recent studies have introduced CNN, LSTM, and Transformer models. There is still research on the use of rarely used multimodal, multilingual, and real-time data. Moreover, there are insufficient XAI methods that impede healthcare trustworthiness and real-world implementation. The study highlights critical gaps in medical testing, model generalization, standard comparisons, and diversity in datasets. Future research aims to develop an understandable, scalable PPD detection model. Upcoming investigations should prioritize large-scale, multi-source datasets, incorporate XAI and improved feature representation, and healthcare-oriented assessments.