Clinical tools to detect Postpartum Depression based on Machine learning and EEG: A Review

Abstract

The expectant mother experiences considerable anatomical and physiological changes during pregnancy that induce stress, and an acute response to stress induces a state of anxiety. When this stress is prolonged, depression may develop insidiously, impairing one's ability to cope with stress. Around the world, up to 15% of new mothers may experience postpartum depression, a mental condition. It has been linked to poor maternal and newborn bonding, reduced breastfeeding initiation, and poor mental health outcomes for mothers, children, and infants. Even though they are common, up to 50% of postpartum issues may go undiagnosed or untreated. Therefore, it is imperative to detect symptoms in the early stages for prevention and intervention. In this paper, we review the existing techniques for PPD detection and prevention and cite their merits and demerits. The secondary questionnaire-based screening tools can only act after the onset of depression but before the development of the disorder, while primary prevention approaches, which use powerful machine learning algorithms, can work before the emergence of symptoms themselves with a higher accuracy. The most accurate classifier has been proven to be Support Vector Machine in the vast majority of instances. Recently, Electroencephalogram (EEG) has received the most research attention because of its superior features in measuring cognitive abilities in real-time. Therefore, adopting EEG-based methodologies could help address inconsistencies caused by earlier approaches. Accordingly, the review suggests exploring the use of EEG for improving the detection of postpartum depression to account for early interventions.