Early Life Stress Detection Using Physiological Signals and Machine Learning Pipelines.

Area of Science:

• Psychoneuroimmunology
• Developmental Psychology
• Computational Biology

Background:

• Pregnancy and early childhood are critical developmental periods with heightened immunological sensitivity.
• Exposure to stress during these times can significantly increase health risks for both mother and child.
• Early Life Stress (ELS) is linked to adverse outcomes in psychological development, metabolism, and cardiovascular health.

Purpose of the Study:

• To investigate the relationship between recalled childhood or pregnancy hardship and inflammatory imbalance in at-risk pregnant women.
• To explore the potential of machine learning for detecting stress using physiological signals during pregnancy.
• To identify strategies for controlling Early Life Stress in pregnant women at risk for future diseases.

Main Methods:

• Retrospective data collection on childhood and pregnancy hardship.
• Assessment of inflammatory imbalance in a cohort of 53 ethnically diverse, low-income women.
• Application of Convolutional Neural Networks (CNNs) for stress detection using short-term physiological data (heart rate, galvanic skin response).

Main Results:

• Analysis of the association between recalled hardship and inflammatory markers.
• Evaluation of CNN performance in identifying stress from physiological signals.
• Identification of specific physiological indicators of stress during pregnancy.

Conclusions:

• Early Life Stress during pregnancy poses significant risks to maternal and child well-being.
• Machine learning approaches show promise for objective stress detection in pregnant populations.
• Further research is needed to develop effective interventions for managing ELS in at-risk pregnant women.