ICASSP2026 – Deep Spatio-Temporal Models for Decoding Purkinje Cell Activity in Tongue Movements

Muhammad Zeeshan, Lorenzo Bina, Laurens WJ Bosman, Chris I De Zeeuw, Muhammad Ali Siddiqi, Murtaza Taj

Abstract:

Decoding cerebellar activity is crucial for understanding the neural basis of motor control and for advancing brain–computer interface (BCI) research. In this work, we investigate whether spike activity from Purkinje cells can be used to classify targeted licking behavior in mice. We propose a hybrid approach utilizing both hand-crafted and convolutional features. We first compute spike count (frequency) and coefficient of variation using inter-spike interval, and create a 1D feature representation. We then design a novel spatio-temporal 2D representation of spike data from multiple electrodes that fuses simple and complex spike information across electrodes, thereby providing a holistic view of cerebellar activity in the form of a 2D map. We then propose two novel architectures, namely PurkinjeSpikeNet-CNN (PSN-CNN) and PurkinjeSpikeNet-LSTM (PSN-LSTM). PSN-CNN performs spatial analysis via convolutional features, whereas PSN-LSTM performs spatio-temporal analysis via recurrent neural network. Both PSN-CNN and PSN-LSTM surpass state-of-the-art accuracy by over $3.58\%$. This demonstrates that the proposed solution is both accurate and possesses generalization capabilities.

Code: The code is available at Github .
PDF: PDF
Poster: PDF

Text Reference:

 Muhammad Zeeshan, Lorenzo Bina, Laurens WJ Bosman, Chris I De Zeeuw, Muhammad Ali Siddiqi, Murtaza Taj, "Deep Spatio-Temporal Models for Decoding Purkinje Cell Activity in Tongue Movements," in Proc. of the IEEE Int. Conf. on Acoustics, Speech, and Signal Processing (ICASSP), 2026

Bibtex Reference:

@inproceedings{localizationlensMICCAI2025,
  author={M.H. Zahid, and M. Taj},
  title={Deep Spatio-Temporal Models for Decoding Purkinje Cell Activity in Tongue Movements},
  booktitle={Proc. of the Int. Conf. on Acoustics, Speech, and Signal Processing (ICASSP)},
  year={2026},
}