Automatic mental health identification method based on natural gait pattern.

Area of Science:

• Biomedical Engineering
• Psychiatry
• Machine Learning

Background:

• Mental health conditions like depression and anxiety are global health challenges, affecting hundreds of millions worldwide.
• Traditional one-to-one consultations face limitations in meeting the vast needs of the sub-healthy population due to resource constraints.
• There is a need for scalable, accessible methods for mental health assessment.

Purpose of the Study:

• To introduce a novel method for mental health recognition by analyzing daily gait patterns.
• To identify individuals with potential symptoms of depression and anxiety using gait data.
• To explore the feasibility of using gait analysis for large-scale mental health screening.

Main Methods:

• Eighty-eight participants' gaits were recorded using digital cameras.
• Participants completed the Patient Health Questionnaire (PHQ-9) for depression and the Generalized Anxiety Disorder Scale (GAD-7) for anxiety.
• 18 key body trunk points were tracked, and time-domain and frequency-domain features were extracted. Machine learning models were developed for recognition.

Main Results:

• The proposed gait-based method demonstrated feasibility and effectiveness in recognizing mental health conditions.
• A medium correlation was achieved for depression recognition (correlation coefficient > 0.5) and anxiety recognition (correlation coefficient > 0.4).
• The models successfully identified clinically significant symptoms of depression and anxiety based on gait.

Conclusions:

• Gait analysis presents a viable, low-cost, and convenient approach for mental health recognition.
• This method can be integrated into daily monitoring systems for mental well-being.
• The approach holds promise for large-scale preliminary screening of mental health conditions.