Stress Inference from Abdominal Sounds using Machine Learning.

Area of Science:

• Biomedical Engineering
• Computational Health
• Psychophysiology

Background:

• Stress is a global health epidemic affecting over a third of the population.
• Chronic stress negatively impacts both physical and mental well-being.
• The gastrointestinal system's connection to stress is increasingly recognized.

Purpose of the Study:

• To develop and evaluate a machine learning methodology for stress detection using abdominal sounds.
• To explore the efficacy of acoustic and mood-related features for stress classification.
• To investigate the feasibility of using wearable technology for continuous stress monitoring.

Main Methods:

• Collected 104 hours of abdominal sounds from eight participants under controlled stress (Stroop test) and relaxation (guided meditation) conditions.
• Utilized a custom wearable device integrated into a belt form-factor for data acquisition.
• Applied traditional machine learning methods, including feature extraction, reduction, and a multilayer perceptron classifier.

Main Results:

• Achieved 77% accuracy in detecting stress exposure through analysis of abdominal sounds.
• Evaluated the impact of acoustic features, mood state features, and domain-specific features on classification performance.
• Demonstrated the potential of combining different feature types for improved stress detection.

Conclusions:

• This feasibility study confirms a link between gastrointestinal activity and stress levels.
• Abdominal sound analysis using machine learning presents a novel and promising approach for stress inference.
• Wearable technology offers a viable pathway for non-invasive, continuous stress monitoring.