A Unified Optimization Model of Feature Extraction and Clustering for Spike Sorting

Abstract

Spike sorting technologies support neuroscientists to access the neural activity with single-neuron or single-actionpotential resolutions. However, conventional spike sorting technologies perform the feature extraction and the clustering separately after the spikes are well detected. It not only induces many redundant processes, but it also yields a lower accuracy and an unstable result especially when noises and/or overlapping spikes exist in the dataset. To address these issues, this paper proposes a unified optimization model integrating the feature extraction and the clustering for spike sorting. Unlike the widely used combination strategies, i.e., performing the principal component analysis (PCA) for spike feature extraction and the K-means (KM) for clustering in sequence, interestingly, this paper finds the solution of the proposed unified model by iteratively performing PCA and KM-like procedures. Subsequently, by embedding the K-means++ strategy in KM-like initializing and a comparison updating rule in the solving process, the proposed model can well handle the noises and overlapping interference as well as enjoy a high accuracy and a low computational complexity. Finally, an automatic spike sorting method is derived after taking the best of the clustering validity indices into the proposed model. The extensive numerical simulation results on both synthetic and realworld datasets confirm that our proposed method outperforms the related state-of-the-art approaches.

A robust spike sorting method based on the joint optimization of linear discrimination analysis and density peaks

“…HC1 is a publicly available in-vivo dataset, which contains the extracellular and intracellular signals from rat hippocampal neurons with silicon probes 44 . It is a widely used benchmark recorded with sparse electrodes 22,31,38 . We used the simultaneous intracellular recording as the label information of extracellular recording to obtain partial ground truth 44 .…”

“…Its operation also calls LDA-Km which brings additional computation complexity. Recently, the concept of joint optimization of feature extraction and clustering has been adopted to construct a unified optimization model of PCA and Km-like procedures 38 , which integrates the feature extraction and clustering steps for spike sorting.…”

A Robust Spike Sorting Method based on the Joint Optimization of Linear Discrimination Analysis and Density Peaks

“…HC1 is a publicly available in-vivo dataset, which contains the extracellular and intracellular signals from rat hippocampal neurons with silicon probes [44]. It is a widely used benchmark for spike sorters oriented sparse electrodes [22, 31, 38]. We used the synchronized intracellular recording as the label information of extracellular recording to obtain partial ground truth [44].…”

From End to End: Gaining, Sorting, and Employing High-Density Neural Single Unit Recordings

“…Discrete derivatives (Zamani et al, 2018 ) or optimal wavelet transforms (Yang and Mason, 2017 ; Soleymankhani and Shalchyan, 2021 ), which are sub-band selective, can stand for filtering as well (Soleymankhani and Shalchyan, 2021 ), whereas zero crossing features (Oh et al, 2017 ) or first and second derivative spike features (Caro-Martín et al, 2018 ) are methods that can tackle this condition. These methods are concentrated on global features gripping waveform morphology similarities of action potentials, but local feature extraction, e.g., Laplacian eigenmaps, could constitute another strategy as well (Chah et al, 2011 ; Huang et al, 2021 ). Regardless of the choice of feature extraction algorithms, by the end of this step, a well-represented feature space should be received, mapping each spike snippet as part of a highly distinguished and densely populated area (Chung et al, 2017 ).…”

“…Hierarchical solutions are mainly represented by Euclidean distance-employing algorithms (Knieling et al, 2017 ), which are the base of optimal filter estimation methods, too (Hassan et al, 2020 ). Graph-based clustering has nearest neighbor interactions at the center, but spectral clustering (Huang et al, 2021 ) or super-paramagnetic clustering in the well-known wave_clus algorithm does also subscribe to this ground (Quian and Nadasdy, 2004 ). Fuzzy-C-means logic considers each action potential as a member of every cluster that has been delineated, and only their affinity degree makes decoding possible (Regalia et al, 2016 ).…”