Data Mining and Machine Learning on the Basis from Reflexive Eye Movements Can Predict Symptom Development in Individual Parkinson’s Patients

Abstract

We are still not in a position to understand most of the brain’s deeper computational properties. As a consequence, we also do not know how brain processes are affected by nerve cell deaths in neurodegenerative diseases (ND). We can register symptoms of ND such as motor and/or mental disorders (dementias) and even provide symptomatic relief, though the structural effects of these are in most cases not yet understood. Fortunately, with early diagnosis there are often many years of disease progression with symptoms that, when they are precisely monitored, may result in improved therapies. In the case of Parkinson’s disease, measurements of eye movements can be diagnostic. In order to better understand their relationship to the underlying disease process, we have performed measurements of reflexive eye movements in Parkinson’s disease (PD) patients. We have compared our measurements and algorithmic diagnoses with experts’ diagnoses. The purpose of our work was to find universal rules, using rough set theory, to classify how condition attributes predict the neurologist’s diagnosis. Prediction of individual UPDRS values only from reflexive saccade (RS) latencies was not possible. But for n = 10 patients, the patient’s age, latency, amplitude, and duration of RS gave a global accuracy in individual patients’ UPRDS predictions of about 80%, based on cross-validation. This demonstrates that broadening the spectrum of physical measurements and applying data mining and machine learning (ML) can lead to a powerful biomarker for symptom progression in Parkinson’s.