Applying machine learning to dissociate between stroke patients and healthy controls using eye movement features...

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Human-Computer Interaction

Background:

• Conventional neuropsychological tests lack ecological validity for real-world scenarios.
• Immersive virtual reality (VR) offers controlled, dynamic simulations for assessment.
• Integrating eye tracking with VR provides detailed behavioral metrics.

Purpose of the Study:

• To evaluate the potential of eye movement analysis in VR for neuropsychological assessment.
• To differentiate between task complexities (short vs. long shopping lists) using eye tracking.
• To distinguish between stroke patients and healthy controls based on eye movement data.

Main Methods:

• Participants (83 stroke patients, 103 controls) navigated a virtual supermarket.
• Eye movements were recorded while participants completed shopping tasks (3 or 7 items).
• Machine learning models (Logistic Regression, Support Vector Machine) analyzed eye tracking data.

Main Results:

• Models predicted shopping list length with an AUC of .76.
• Models differentiated stroke patients from controls with an AUC of .64.
• Revisiting aisles was the most significant feature for both classification tasks.

Conclusions:

• Eye movement data from VR simulations reveal cognitive signatures.
• This approach shows promise for detecting cognitive deficits and clinical applications.
• VR-based eye tracking enhances neuropsychological assessment capabilities.