Can Saccade and Vergence Properties Discriminate Stroke Survivors from Individuals with Other Pathologies? A Machine...

Area of Science:

• Ophthalmology and Neurology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Stroke often causes visual deficits and eye movement disorders.
• Machine learning (ML) has shown potential in identifying various neurological conditions through eye movement analysis.
• REMOBI technology and Pupil Core eye trackers provide precise saccade and vergence measurements.

Purpose of the Study:

• To investigate the efficacy of ML in identifying stroke patients using saccade and vergence eye movement data.
• To compare the performance of different ML models in classifying stroke patients against control groups.
• To assess the potential of this technology as a clinical tool for stroke patient evaluation.

Main Methods:

• Eye movement data (saccades and vergence) were collected using REMOBI technology V3 and Pupil Core.
• Data were analyzed using AIDEAL V3 (Artificial Intelligence Eye Movement Analysis) software, extracting parameters like latency, accuracy, and velocity.
• Three ML models (logistic regression, support vector machine, random forest) were applied to classify stroke patients.

Main Results:

• ML classifiers achieved macro F1 scores up to 75.9% in identifying stroke patients.
• Age-matched analysis further improved classifier performance, particularly when comparing against healthy individuals.
• The study demonstrated the sensitivity of ML-based eye movement analysis for detecting stroke sequelae.

Conclusions:

• ML applied to saccade and vergence parameters is a sensitive method for detecting stroke-related sequelae.
• The combination of REMOBI and AIDEAL technology offers a viable approach for stroke patient identification.
• This technology holds potential for clinical applications in evaluating stroke recovery and neurological deficit progression.