Mental Status Detection for Schizophrenia Patients via Deep Visual Perception

Area of Science:

• Computer Science
• Artificial Intelligence
• Psychiatry

Background:

• Schizophrenia is a mental disorder with significant social impact, often correlated with emotional status and depression.
• Existing methods for emotional status detection primarily rely on facial analysis, potentially limiting comprehensive assessment.
• Accurate mental status detection is crucial for providing effective assessment tools for mental health professionals.

Purpose of the Study:

• To design and develop a novel mental status detection system for schizophrenia patients.
• To provide an advanced assessment tool for mental health professionals to monitor patient conditions.
• To improve the accuracy of inferring mental states, including emotion and depression severity, in individuals with schizophrenia.

Main Methods:

• A multi-task learning framework was proposed for inferring emotion and depression severity.
• A Cross-Modality Graph Convolutional Network (CMGCN) was employed to integrate visual features from face and context.
• Task-aware objective functions and an Emotion Passer module were designed to enhance multi-task learning and knowledge transfer.

Main Results:

• The proposed CMGCN framework significantly improved upon state-of-the-art methods in benchmark dataset experiments.
• The system achieved a mean Average Precision (mAP) of 69.52 in a real-world trial with schizophrenia patients.
• The integration of multi-modal visual features and task-aware learning demonstrated superior performance.

Conclusions:

• The developed system offers a promising approach for objective mental status assessment in schizophrenia.
• The multi-task learning framework with CMGCN effectively captures complex emotional and depression indicators.
• This technology has the potential to aid mental health professionals in diagnosing and managing schizophrenia more effectively.