A new approach for detecting and analyzing cutaneous reflexes during locomotion

During human walking, due to their small amplitude, individual cutaneous reflex responses are difficult to detect in surface EMG recordings. In this study, we present a new algorithm to automatically detect individual cutaneous reflex responses and to extract their corresponding onset latency, amplitude, duration, and sign. To discriminate reflex responses from the intrinsic variability of the background EMG, each stimulated cycle is compared with 10 adjacent nonstimulated cycles, looking for consistent differences. In the first 200 ms after stimulation, reflex responses are detected when ≥ 9/10 of these differences are either positive or negative. This approach does not require amplitude thresholds or fixed time windows for reflex detection. To reduce false detections, a postprocessing step selects 50 nonstimulated cycles randomly, processes them through the algorithm as stimulated cycles, and establishes a minimal reflex duration criterion that it then used to validate the detected responses. Validated responses from an entire test session are then reported on a colormap (reflex activity map) from which specific responses can be identified and quantified. The new method was validated in ten participants, three cutaneous nerves, and two protocols (phase modulation and recruitment curves). Compared with the classical method, the new algorithm showed better performance in terms of detection accuracy, specificity, and reliability. Although tested here to evaluate cutaneous reflexes during human walking, the simplicity of this method is such that it could easily be used with other reflexes, signals, and preparations.