A Novel Spiking Neural Network Approach for Hand Motion Assessment in Awake Neurosurgery

In Awake Neurosurgery (AN), motor functions are traditionally assessed qualitatively by neuropsychologists, relying on visual observation during surgery. This study introduces a non-invasive approach using a Spiking Neural Network (SNN) and the Leap Motion Controller (LMC) to automate the recognition and evaluation of hand motion tasks, supporting medical specialists in preserving eloquent brain areas and maintaining patients' postoperative functional autonomy. The proposed SNN processes state-of-the-art features extracted from a custom LMC dataset, achieving a mean accuracy of 68.1% in 10-fold cross-validation. An ablation study on population coding (1 to 100 neurons per class). Compared to Long Short-Term Memory (LSTM) models, the SNN offers greater stability across folds, despite slightly lower accuracy. This lightweight, efficient approach demonstrates potential as a decision-support tool for AN.