Action Potential Features: Computation and Spike Sorting of Human C-Nociceptor Action Potentials as obtained via Microneurography Recordings

Abstract

Spike sorting represents a persistent challenge in electrophysiology, particularly in extracellular nerve recordings containing signals from several nerve fibers. This issue is exacerbated in microneurography recordings from peripheral unmyelinated afferents in awake humans, which are responsible for pain sensation. This is due to the similarity of spike shapes originating from different fibers, low signal-to-noise ratios, and shape-distorting overlaying signals. Here, we present the first systematic assessment of morphology-based spike sorting in multiple recordings from two microneurography laboratories. We created dedicated ground truth datasets by employing semi-manual labelling methods enabling the comparison of supervised and unsupervised sorting methods for different feature sets. A strong advantage of the supervised approach was observed, while no single feature set showed a global advantage. Further, the high diversity of the results was linked to the per-recording fiber number and spike morphologies. To extend this first systematic assessment of the spike sorting problem in microneurography, our open-source pipeline enables reproducible sortability analysis of any extracellular recordings of neuronal activity if electrical stimulation of the nerve fibers is possible. The achieved advancement of spike sorting for microneurography lays the foundation for gaining insights into the neural coding of pain and itch signals in a clinical context.