Analyzing Generative AI and Machine Learning in Auto-Assessing Schizophrenia's Negative Symptoms

Area of Science:

• Psychiatry and Computational Neuroscience
• Artificial Intelligence in Healthcare
• Clinical Assessment Tools

Background:

• Traditional schizophrenia negative symptom assessments are subjective and time-consuming.
• Objective and efficient evaluation methods are needed for schizophrenia.
• Negative symptoms include expression (EXP) and motivation and pleasure (MAP) domains.

Purpose of the Study:

• To evaluate an automated system using generative AI (GenAI) and machine learning (ML) for schizophrenia negative symptom assessment.
• To determine the efficacy of AI and ML in assessing expression (EXP) and motivation and pleasure (MAP) domains.
• To provide more objective and efficient clinical evaluations for schizophrenia.

Main Methods:

• Utilized a semi-structured interview protocol based on the Clinical Assessment Interview for Negative Symptoms.
• Conducted audio- and video-recorded interviews with schizophrenia patients by a senior psychiatrist.
• Employed ML for visual/audio data analysis of EXP and GenAI for transcript analysis of MAP.

Main Results:

• ML-based EXP assessment showed moderate to substantial reliability (ICC3,1=0.65, kappa=0.62).
• GenAI-based MAP assessment demonstrated good reliability (ICC3,1=0.82, kappa=0.77).
• The system exhibited strong correlations with clinician ratings (Pearson's r ≥ 0.54) and low error rates.

Conclusions:

• GenAI and ML are effective for automated assessment of schizophrenia negative symptoms.
• The developed system enhances consistency and efficiency in clinical evaluations.
• This technology holds potential for improving schizophrenia diagnosis and management.