A new framework for mental illnesses diagnosis using wearable devices aided by improved convolutional neural network

Area of Science:

• Digital Health
• Artificial Intelligence in Medicine
• Psychiatry

Background:

• Mental Health Disorders (MHDs) are a global concern, often exacerbated by modern life stressors.
• Early diagnosis and treatment are crucial for managing MHDs, yet distinguishing between them is challenging due to varied symptoms.
• Wearable technology, like smart watches, offers a promising avenue for continuous bioactivity monitoring.

Purpose of the Study:

• To present a novel framework for diagnosing mental illnesses using smart watch motor activity data.
• To analyze complex behavioral patterns associated with conditions such as schizophrenia and depression.
• To improve the accuracy and efficiency of mental health disorder detection.

Main Methods:

• Utilizing a framework that analyzes motor activity data from smart watches.
• Encoding time-series behavioral data into image patterns via a modified Markov Transition Field.
• Employing a modified Convolutional Neural Network with attention pooling for patient classification.

Main Results:

• Achieved high diagnostic accuracy: 96.6% for schizophrenia and 94.85% for depression.
• Demonstrated strong precision: 92.1% for schizophrenia and 96.77% for depression.
• The system effectively analyzes individual motor activity to diagnose mental illness.

Conclusions:

• The proposed system demonstrates superior performance compared to existing methods for detecting mental health disorders.
• Smart watch data analysis offers a viable, non-invasive approach for diagnosing conditions like schizophrenia and depression.
• This AI-driven framework holds significant potential for advancing digital mental healthcare.