Exploratory eye movement characteristics to aid screening of schizophrenia and bipolar disorder: A cross-sectional...

Current psychiatric diagnoses lack objective criteria, and this study aims to evaluate EEM as a potential tool for improving diagnostic objectivity across psychiatric disorders. The Exploratory Eye Movement (EEM) paradigm was used to analyze eye-movement data from patients with schizophrenia or bipolar disorder and healthy controls. Key metrics included Number of Eye Fixations (NEF), Total Eye Scanning Length (TESL), Mean Eye Scanning Length (MESL), Cognitive Search Score (CSS), and Responsive Search Score (RSS). Group differences were examined with ANOVA and effect sizes, and diagnostic performance was assessed using ROC curves. Feature importance and classification were evaluated with machine learning models using 10-fold cross-validation. Significant age differences were noted between groups, potentially influencing feature selection. NEF and RSS were identified as the most discriminative features, particularly in schizophrenia vs. healthy controls (Cohen's d = -0.79 and -1.12). ROC analysis showed RSS (AUC = 0.84) and NEF (AUC = 0.78) as the top indicators. The SVC model, incorporating demographic features and the top two MI-selected eye movement features (NEF, RSS), achieved an AUC of 0.80 and an F1 score of 0.60, outperforming other models. EEM-based indicators such as RSS, NEF can serve as an adjunctive screening tool to help diagnose but not a stand-alone diagnostic method. Implementing standardized EEM examination procedures in clinical practice is potentially valuable for the early screening of these conditions. Future research could explore integrating EEM with other diagnostic methods to construct an intelligent and comprehensive assessment system.