Frequency-Based Filtering for Automated Spike Detection in Human Microneurography

Dominika, Darabos and Alina, Troglio and Andrea, Fiebig and Anna, Maxion and Ekaterina, Kutafina and Barbara, Namer and Bognár, Gergő and Kovács, Péter (2026) Frequency-Based Filtering for Automated Spike Detection in Human Microneurography. In: Proceedings of the Workshop Biosignale 2026. Innsbruck University Press (IUP), Innsbruck, pp. 23-25. ISBN 9783991061946

Preview

Text Biosignale_2026_DD.pdf - Published Version Download (2MB) | Preview

Abstract

Although microneurography is a powerful tool for recording activity from individual human peripheral nerve fibers, its broader adaptation is still limited due to the characteristics of the data, which prevent most existing spike detection algorithms from being used effectively. Since these are extracellular recordings collected with microelectrodes, they contain a significant amount of noise, resulting in a low signal-to-noise ratio. This noise can easily mask the spike-shaped responses of nerve fibers, whose shape can also vary over time, making it difficult to develop detectors that generalize reliably across different spike shapes. In this study, we analyze the recordings in the frequency domain and identify frequency components that can be removed with minimal compromise to the underlying neural activity. Using the weighted Hermite Variable Projection Neural Network applied to sliding-window segments of the signal, we compare classification performance on the original, low-pass filtered, and band-pass filtered data. The results identify the frequency ranges that carry the most informative components of the neural signal and demonstrate that appropriate filtering can reduce the number of false alarms to 22% of the level obtained with unfiltered data, while only modestly affecting true detections.

Actions (login required)

View Item