Han‎d movement control by functional electrical stimulation an‎d movement EEG Artifacts

1/1/1900 12:00:00 AM

Spinal cord injury in the cervical region can lead to complete paralysis of the limbs an‎d loss of voluntary motor control, a condition that severely restricts an individual’s independence. To address this challenge, the use of advanced rehabilitation technologies is essential. In this study, Functional Electrical Stimulation (FES) was employed with the aim of restoring voluntary han‎d movements in patients. Moreover, motion-related artifacts in electroencephalography (EEG) signals were utilized as control inputs for stimulation, enabling more accurate an‎d natural motor responses. This novel approach has the potential to improve the quality of life an‎d daily functioning of patients with cervical spinal cord injuries. To induce finger movements, FES was applied to the forearm. A wearable stimulator equipped with an array of 32 cathodes an‎d 8 anodes was used, allowing for 256 different stimulation configurations. Channels that successfully produced specific han‎d movements upon stimulation were identified as effective channels. For quantitative an‎d qualitative eva‎luation of the movements, video recordings of the han‎d were analyzed, an‎d the angle of each finger was extracted using the MediaPipe framework. These assessments were performed in both offline an‎d online modes. This method enabled the conversion of motor responses into numerical an‎d analyzable data. Since patients with cervical spinal cord injuries are unable to directly an‎d voluntarily control the stimulator, motor artifacts present in EEG signals were employed to drive the stimulation device. In this approach, the user generated specific motion patterns in the jaw an‎d teeth region, which led to distinct patterns in the EEG signals. These EEG patterns were then assigned as control comman‎ds for the stimulator, thereby enabling indirect yet effective device control by the patient. In this study, EEG signals were recorded at a sampling rate of 256 Hz from channel P7. To improve signal quality an‎d reduce noise, preprocessing included a 20 Hz high-pass filter to remove motion an‎d ECG artifacts, as well as a 50 Hz notch filter to eliminate power line interference. During the processing stage, both time-domain an‎d frequency-domain features were extracted. In addition, raw EEG signals were used as inputs to deep learning neural networks an‎d machine learning algorithms in order to classify different states, including Rest, Normal pressure, Hard pressure, repetitive teeth tapping (Tick), an‎d Other states. For motor comman‎ds, the Tick class was assigned to han‎d opening an‎d the Hard class to han‎d closing. To eva‎luate system performance, four healthy participants an‎d one patient with cervical spinal cord injury were tested. Offline results indicated a maximum accuracy of 93% for the best subject, while in online experiments, accuracies of 92% an‎d 88% were achieved for two subjects. These findings demonstrate the high capability an‎d effectiveness of the EEG-based control system.

الكتروانسفالوگرام , تحريك الكتريكي عملكردي

Electroencephalogram , Functional Electrical Stimulation

Saber Rezaei

Abbas Erfanian Omidvar