Automatic Screening of Parkinson's Disease from Visual Explorations

2025-09-01

AI Summary

Conventional oculomotor biomarkers exhibit limited discriminative power for non-invasive early screening of Parkinson’s disease (PD). Method: This study proposes a novel, vision-based PD screening framework leveraging eye-tracking features during visual exploration tasks. To overcome the limitations of traditional metrics, we innovatively integrate low-level oculomotor features (e.g., saccade frequency, fixation duration) with high-level gaze clustering region features, and design a mixture-of-experts ensemble model tailored for multi-task paradigms (e.g., free viewing, target search) and binocular coordination analysis. Contribution/Results: The framework effectively fuses heterogeneous eye-tracking data, enhancing inter-individual discriminability and robustness. On an independent test set, it achieves an AUC of 0.95—significantly outperforming single-feature or monocular modeling approaches. Results demonstrate that visual exploration paradigms combined with multi-granularity oculomotor modeling constitute a highly accurate, non-invasive pathway for PD screening.

Technology Category

Computer Vision: Multi-modal VisionSearch and Optimization: Mixed Discrete/Continuous SearchIntelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

User Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingEconomics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAISearch and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Eye movements can reveal early signs of neurodegeneration, including those associated with Parkinson's Disease (PD). This work investigates the utility of a set of gaze-based features for the automatic screening of PD from different visual exploration tasks. For this purpose, a novel methodology is introduced, combining classic fixation/saccade oculomotor features (e.g., saccade count, fixation duration, scanned area) with features derived from gaze clusters (i.e., regions with a considerable accumulation of fixations). These features are automatically extracted from six exploration tests and evaluated using different machine learning classifiers. A Mixture of Experts ensemble is used to integrate outputs across tests and both eyes. Results show that ensemble models outperform individual classifiers, achieving an Area Under the Receiving Operating Characteristic Curve (AUC) of 0.95 on a held-out test set. The findings support visual exploration as a non-invasive tool for early automatic screening of PD.

Problem

• Automatically screen Parkinson's Disease using eye movements
• Develop gaze-based features from visual exploration tasks
• Create ensemble models for improved diagnostic accuracy

Innovation

• Gaze-based features from visual exploration tasks
• Combining classic oculomotor with gaze cluster features
• Mixture of Experts ensemble integrating multi-test data