Major features for bradykinesia classification in Parkinson diseased patients

Monitoring motor function in patients with Parkinson's disease is essential to improve symptom administration and avoid pathological complications. The goal of this project is to examine the best possible features from bradykinesia patient movements taken from wearable sensors installed on their limbs. Elimination of constant features, correlation-based feature selection, and one-way ANDVA were the methods used to find important features used for the training of machine learners from a combination including temporal, spectral, and statistical features. The results show that achieving higher machine learner accuracies is ensured by critically important features that include the fast Fourier transform, maximal amplitude, Benford correlation, and sample entropy.