Objective: Eye tracking has been suggested as a source of prospective biomarkers in Parkinson’s Disease (PD), and recent work has demonstrated the use of machine learning to classify PD and its cognitive spectrum based on oculomotor features extracted during an antisaccade task. We extend that work here, demonstrating increased sensitivity in an unstructured, free-viewing task to both PD and Progressive Supranuclear Palsy (PSP).

Background: Neurodegenerative Disorders (NDDs) are the leading and increasing cause of disability worldwide. Increased rates of PD are particularly alarming, and there is an urgent need for sensitive and specific biomarkers that can differentially predict PD, especially in early and prodromal phases of the disease, such that a personalized medicine approach to therapeutics can be planned and tracked longitudinally. Constructing objective measures of PD is complicated by its cognitive and motor spectrum of disorder, as well as the presence of atypical disorders such as PSP.

Method: 120 PD participants, 8 PSP participants, and 97 age-matched control participants without neurological dysfunction performed a naturalistic free-viewing task while their eyes were tracked at high fidelity. From 10 minutes of movies, we extracted measures of saccade, pupil, and blink. These measures were used to train a support vector machine classifier. The classifier was tuned, and performance was measured by the area under receiver operating characteristic curves (ROC-AUC) through a held-out test set and cross validation.

Results: The average ROC-AUC on the test set was 0.89, with 0.95 for PSP when compared to the other groups. The average ROC-AUC via cross-validation was 0.867 (95% CI: [0.799 0.935]). See Fig. 1.

Conclusion: PD and PSP can be predicted with high sensitivity using a free-viewing eye-tracking paradigm. ROC-AUC was comparable and superior to that of the antisaccade task. An unstructured task is easy to administer, can be used on all age-ranges and spectrums of cognition, and it is fast and non-invasive. Our next steps are to assess the performance of this classifier on a naïve test-set from an independent site.

« Back to 2023 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/machine-learning-classifies-parkinsons-disease-and-progressive-supranuclear-palsy-on-saccade-pupil-and-blink-measures-during-a-naturalistic-free-viewing-task/