Machine Learning on the Video Basis of Slow Pursuit Eye Movements Can Predict Symptom Development in Parkinson’s Patients

Abstract

We still do not know exactly how brain processes are affected by nerve cell deaths in neurodegenerative diseases such as Parkinson’s (PD). Early diagnosis when symptom progressions are precisely monitored may result in improved therapies. In the case of PD, measurements of eye movements (EM) can be diagnostic. In order to better understand their relationship to the underlying disease process, we have performed measurements of slow (POM) eye movements in PD patients. We have compared our measurements and algorithmic diagnoses with doctor’s diagnoses. We have used rough set theory and machine learning (ML), to classify how condition attributes predict the neurologist’s diagnosis. We have measured pursuit ocular movements (POM) for three different frequencies and estimated patients’ performance by gain and accuracy for each frequency. We have tested ten PD patients in four sessions related to combination of medication and DBS treatments. We have obtained a global accuracy in individual patients’ UPRDS III predictions of about 80%, based on cross-validation. This demonstrates that POM may be a good biomarker helping to estimate PD symptoms in automatic, objective and doctor-independent way.