Fuzzy Logic as a Control Strategy to Command a Deep Brain Stimulator in Patients with Parkinson Disease

Abstract

Deep brain stimulation (DBS) of the subthalamic nuclei (STN) is the most used surgical treatment to improve motor skills in patients with Parkinson’s disease (PD) who do not respond well to pharmacological treatment. Currently, DBS operates in open-loop mode stimulating with constant parameters. A way to improve the therapy is modeling a closed-loop DBS system that automatically adjusts the stimulation parameters based on the clinical and/or neural state of the patient. Fuzzy logic was used to design three models: two Mamdani-type models, varying the defuzzification method, and a Sugeno-type one. The inputs for all models were the beta oscillation power calculated from local field potential and the magnitude of the acceleration recorded from a smart watch. Total electrical energy delivered (TEED) and theoretical charge density in the STN (Q) was calculated in order to evaluate the performance of the algorithms. The model with the lowest TEED was the Mamdani-Bisector type with 136.80 ± 59.07 μW, followed by the Mamdani-Centroid type with 138.90 ± 57.22 μW, being 66% lower than open loop DBS. Mamdani-Centroid had the lowest maximum Q and a smooth output for the trials tested, with gradual changes in the control surface. Therefore, the Mamdani-Centroid model resulted to be the best control strategy to implement a closed-loop DBS system in our study.