Predicting Motor Intent from Residual Neck Muscle Activity in Individuals with Neck Weakness from ALS

The long-term goal of this work is to restore dexterous and intuitive head-neck motion to patients with Amyotrophic Lateral Sclerosis (ALS). ALS is an idiopathic disease characterized by progressive paralysis. Some patients experience neck weakness such that their heads permanently drop to their chests, causing pain and extreme difficulty eating, navigating, and socializing. We previously developed the Utah Neck Exoskeleton, a powered neck brace that supports the head and uses electric motors to move the head in a large range of motion, counteracting head drop. However, the exoskeleton has been controlled either with a joystick or gaze tracking, both of which are difficult to use for parts of the ALS population. Here, we show that the residual neck muscles of ALS patients with neck weakness can be used to determine intended neck position and motion. Electromyographic (EMG) signals were recorded from the neck muscles of two individuals with ALS, low clinical functional scores, and self-reported neck weakness. EMG was then mapped to either steady-state head position or the direction of head motion using convolutional neural networks. Despite the patients having neck weakness and limited range of motion, EMG signals were sufficient to accurately classify both steady-state head position and the direction of head motion (97.1% and 83.12% median accuracy, respectively). As such, this work demonstrates that EMG may serve as a dexterous and intuitive control modality for real-time head-neck movement, and in conjunction with the Utah Neck Exoskeleton, may ultimately improve quality of life for individuals with head drop.Clinical RelevanceResidual neck muscle activity in ALS patients can be recorded via surface EMG and potentially used to reliably predict intended head position and motion.