Robust Spike Sorting Using Dual Tree Complex Wavelet Transform: Overcoming Traditional Limitations

Abstract

Accurate spike sorting is vital for understanding the neural network dynamics of the brain through extracellular neural recordings. Traditional feature extraction methods like the Haar wavelet and principal component analysis (PCA) face limitations in high-noise environments and when dealing with similar spike waveforms. This study proposes the use of the dual-tree complex wavelet transform (DT-CWT), which is an enhanced version of the traditional discrete wavelet transform, as a feature extractor in spike sorting. The proposed approach aims to leverage the approximate shift invariance and superior time-frequency localization capabilities of the DT-CWT in spike discrimination. A simulated dataset with varying noise levels and spike waveform similarities was used for validation. The performance of DT-CWT was compared with Haar wavelet and PCA using both original and modified superparamagnetic clustering approaches. Results demonstrated that the DT-CWT-based feature extraction achieved significantly lower average error rates across all subsets, especially in challenging conditions with high noise and similar waveforms. Moreover, the DT-CWT's computational efficiency and suitability for real-time, on-chip implementations in next-generation wireless brain computer interface devices highlight its practical advantages. By providing robust and discriminative features under stringent resource constraints, the DT-CWT-based approach outperforms traditional methods, establishing itself as a preferred choice for scalable and efficient spike sorting in demanding scenarios.