A new algorithm for drift compensation in multi-unit recordings of action potentials in peripheral autonomic nerves over time

Abstract

BackgroundPeripheral autonomic nerves control visceral organs and convey information regarding their functional states and are, therefore, potential targets for new therapeutic and diagnostic approaches. Conventionally recorded multi-unit nerve activity in vivo undergoes slow differential drift of signal and noise amplitudes, making accurate monitoring of nerve activity for more than tens of minutes problematic. New MethodWe describe an on-line drift compensation algorithm that utilizes recursive least-squares to estimate the relative change in spike amplitude due to changes in the nerve-electrode interface over time. ResultsWe tested and refined our approach using simulated data and in vivo recordings from nerves supplying the small intestine under control conditions and in response to gut inflammation over several hours. The algorithm is robust to changes in recording conditions and signal-to-noise ratio and applicable to both single and multi-unit recordings. In uncompensated records, drift prevented “spike families” and single units from being discriminated accurately over hours. After rescaling, these were successfully tracked throughout recordings (up to 3 h). Comparison with existing MethodsExisting methods are subjective or compensate for drift using spatial information and spike shape data which is not practical in multi-unit peripheral nerve recordings. In contrast, this method is objective and applicable to data from a single differential multi-unit recording. In comparisons using simulated data the algorithm performed as well as or better than existing methods. ConclusionsResults suggest our drift compensation algorithm is widely applicable and robust, though conservative, when differentiating prolonged responses from drift in signal. Extracellular nerve recordings; drift compensation; chronic nerve recordings; closed-loop; multi-unit activity; spike discrimination; recursive least squares; real-time