On optimization of an embedded ATC-FES system for synergic muscles actions execution

The Functional Electrical Stimulation (FES) exploits low-energy electrical pulses to retrain, or even restore, the functional mobility in patient affected by neuromuscular disorders. The FES effectiveness can be increased modulating the stimulation delivery using the activation pattern of muscles. The ATC is an event-driven processing technique which can be applied to the Surface ElectroMyoGraphy (sEMG) to estimate the muscle contraction force with a low-power approach, representing an effective solution for controlling the FES therapy. The aim of this project is the optimization of an embedded ATC-controlled FES system, which was developed using Python programming language. The application of the system relies on two calibration phases: the first calibration determines the maximum ATC value expressed during the muscle contraction, whereas in the second phase stimulation parameters are tuned to induce the execution of functional movements. In the past version of the system, the two calibrations could not manage more than a single channel at a time and the stimulation profile used for setting FES parameter did not represent the physiological muscle activation. In this project the Profile Extraction algorithm (PE) has been proposed to extent the calibration to functional movements based on the synergic activation of multiple muscles. Specifically, the PE allows the system to extract a multichannel ATC profile from the repetitive execution of a specific movement, representing a statistical information of the voluntary muscle activity. The resulting activation profile can be used to calibrate the maximum ATC value of each channel and to produce a biomimetic stimulation of the patient even during the FES parameters calibration. Moreover, extracted profiles can be stored and used to deliver fully automated FES therapies. In the first part of the project the PE algorithm has been developed using Python programming language: the processing pipeline receives the sequence of ATC values from each acquisition channel and computes the segmentation of movements in real-time. When the acquisition is terminated, segmented movements undergo the final processing phase, which firstly selects the most correlated movements and secondarily aligns selected data maximizing their correlation. Statistical ATC profiles are eventually extracted for each acquisition channel, performing the median among aligned movements. In the second part the ATC-controlled FES system is applied to the stimulation of a multichannel functional task, named drinking task. Initially, a feasibility study has been conducted breaking down the task in the rotation of the glenohumeral joint and the elbow joint, controlled by the contraction of anterior deltoid (AD) and biceps brachii (BB) respectively. An experimental protocol has been designed to validate the multichannel application of the ATC-FES system: eight therapist-patient couples are involved in structured experimental sessions, in which movement trajectories are collected using a motion tracking system. Sessions are performed testing different functional tasks (single channel, multichannel) and stimulation types (biomimetic profile, general pyramidal pattern) delivered during FES calibration. For each couple the validation is performed comparing movement trajectories in different experimental conditions.