Visual Search Processing Patterns in Patients with Depressive Disorder: Attentional Evidence Based on Eye-Tracking Technology

Background Depressive disorder (DD) is associated with attentional dysfunction, but objective measures of altered visual-attention processing remain limited. We characterized eye-tracking patterns during visual search and evaluated their potential adjunctive diagnostic value. Methods 42 patients with DD and 43 healthy controls (HC) performed a visual-search task with eye tracking. Forty-two features derived from six areas of interest (AOIs) and seven eye-tracking metrics were compared using false-discovery-rate (FDR) correction. Associations with scores on the 24-item Hamilton Depression Rating Scale (HAMD-24) were examined within the DD group. Predictive performance was evaluated using repeated stratified nested five-fold cross-validation. Results The final sample included 40 DD and 43 HC participants. The DD group showed longer reaction times than HC(median, 1546.73 vs 1241.52; P =0.004), whereas participant-level accuracy did not differ significantly (median, 1.00 vs 0.99; P =0.775). Twenty-five eye-tracking features showed FDR-corrected between-group differences, but none was significantly associated with HAMD-24 scores after correction. Four features were selected in at least 80% of the 50 outer-training models: fixation count in AOI-C, gaze duration in AOI-E, fixation count in AOI-D, and saccadic amplitude in AOI-D. Across ten repetitions, the mean out-of-fold AUC was 0.990 (SD=0.007; range, 0.971-0.995), with mean accuracy of 95.7%, sensitivity of 94.4%, specificity of 96.9%, and a Brier score of 0.052. The ensemble AUC based on averaged out-of-fold probabilities was 0.996 (95% CI, 0.988-1.000). Conclusions DD was associated with distinct visual-attention and eye-movement patterns. Eye-tracking features may represent candidate markers with potential adjunctive diagnostic value, but external validation in larger independent cohorts is required.