EyeMap: A fusion-based method for eye movement-based visual attention maps as predictive markers of parkinsonism

Akshay S ¹, Amudha J ², Amitabh Bhattacharya ³

⁰Department of Computer Science, School of Computing, Amrita Vishwa Vidyapeetham, Mysuru, Karnataka, India.

Methodsx +

|

September 25, 2025

View abstract on PubMed

Summary

EyeMap visualizes eye movement patterns for Parkinson's disease (PD) detection. This AI-driven method analyzes gaze data, identifying PD-specific anomalies without manual feature engineering.

Area Of Science

Ophthalmology
Neuroscience
Computer Science

Background

Eye movement analysis is crucial for understanding neurological disorders.
Current methods may lack interpretability or require manual feature engineering.
Parkinson's disease (PD) often presents with characteristic gaze abnormalities.

Purpose Of The Study

To develop and validate EyeMap, a novel method for visualizing and classifying eye movement patterns.
To enhance diagnostic interpretability of gaze data for neurological conditions like PD.
To enable the detection of PD-specific gaze anomalies using machine learning.

Main Methods

EyeMap employs scanpaths, fixation heatmaps, and gridded Areas of Interest (AOIs) for visualization.
Late-fusion technique combines predictions from modality-specific machine learning and deep learning models.
A new dataset of eye-tracking data from PD patients and healthy controls was generated.

Main Results

EyeMap successfully visualizes and classifies eye movement patterns.
The method enhances diagnostic interpretability by integrating spatial, temporal, and regional gaze elements.
Vision-driven models utilizing EyeMap detected PD-specific gaze anomalies without manual feature engineering.

Conclusions

EyeMap provides a reproducible and adaptable framework for gaze-based analysis.
The method demonstrates the potential of AI in detecting neurological conditions through eye-tracking.
EyeMap offers complementary perspectives on gaze behavior for improved diagnostic insights.

Keywords:

Classification Deep learning Eye tracking Gaze visualization Human-computer interaction Machine learning Visual attention