Assessing Mothers' Postpartum Depression From Their Infants' Cry Vocalizations

Area of Science:

• Neuroscience
• Developmental Psychology
• Computational Biology

Background:

• Postpartum depression (PPD) affects up to 15% of mothers, impacting infant care and potentially leading to infanticide.
• Current PPD diagnosis relies on self-report, susceptible to social desirability bias, potentially underestimating prevalence.
• Infant vocalizations show acoustic differences linked to maternal depression, suggesting a biological pathway.

Purpose of the Study:

• To investigate the potential of using infant cry acoustics to identify maternal postpartum depression (PPD).
• To develop and evaluate a cloud-based machine learning model for PPD detection using infant vocalization features.

Main Methods:

• Acoustic features (fundamental frequency, formants, intensity) were extracted from infant cry recordings (N=56).
• Infants belonged to mothers diagnosed with PPD (N=29) or without PPD (N=27).
• Cloud-based artificial intelligence models were trained to classify maternal depression status based on acoustic features.

Main Results:

• Machine learning models successfully identified maternal PPD from infant cry acoustics.
• The trained model achieved a high accuracy rate of 89.5% in detecting postpartum depression.
• Acoustic properties of infant cries serve as reliable indicators of maternal PPD.

Conclusions:

• Infant vocalizations offer a promising, objective biomarker for identifying maternal postpartum depression.
• Machine learning analysis of infant cry acoustics presents a novel, accurate method for PPD screening.
• This approach may overcome limitations of self-report measures in PPD diagnosis.