Search Algorithm Optimization for a Neuromuscular Electrical Stimulation Device : A Stochastic Approach for Electrode Pair Selection

Sammanfattning

Neuromuscular electrical stimulation (NMES) has gained prominence as a prevention method for deep vein thrombosis by inducing muscle contractions through electrical impulses. A key factor in achieving effective stimulation, while ensuring patient compliance, is the accurate localization of motor points. However, identifying these points remains a significant challenge. Matrix Muscle Support AB is currently developing an NMES system that automates this process using a search algorithm. The existing baseline algorithm relies on exhaustive search, which could become increasingly inefficient as the number of electrode combinations grows, indicating the need for alternative approaches. In this project, a novel stochastic-based search algorithm, designed to optimize electrode pair selection in a multichannel NMES system, was investigated. The algorithm applies random perturbations during stimulation and continuously evaluates electrode performance based on motion feedback from flex sensors. Implemented through a React-based front end and Bluetooth-enabled interface, the system enables real-time optimization of current intensity and electrode ranking. Performance evaluation confirmed the reliability of the algorithm, showing consistent identification of the highest-ranked electrodes, especially at ranks 1 and 2, across different trials and subjects. The best-performing electrodes also demonstrated high consistency, with structural similarity index (SSIM) values above 0.75 when comparing electrode-response heatmaps across repeated runs. These results provide preliminary evidence that stochastic optimization can be a promising alternative to exhaustive search, but further hardware integration and testing are needed to validate the algorithm’s potential for broader clinical application.