Early Detection of Depression and Treatment Response Prediction using Machine Learning: A Review

Abstract

The turn of the century saw tremendous development in using Machine Learning (ML) for medical diagnosis. Majority of the prediction models or the diagnostic models were based on imaging. These methods extensively rely on huge dataset to develop a deep learning model to effectively diagnose conditions like Alzheimer's, Arrythmia, Lung Cancer, Breast Cancer among others. Early detection of depression is a major hurdle in effective and timely treatment. This paper explores the existing corpora of work on application of ML for early detection of depression. Further, a review on treatment response prediction is also elucidated. Relatively, the apparent dearth of work in depression has been found to be due to the scanty imaging dataset available on depression. We have reviewed the existing body of work on depression that have used imaging biomarkers, mobility traces, multi-modal fusion of audio, video and text. Our hope in writing this paper is to spur and catalyze research in early detection of depression and effective prediction of response to its treatment.