Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

E-Sort: Empowering End-to-end Neural Network for Multi-channel Spike Sorting with Transfer Learning and Fast Post-processing

Yuntao Han, Shiwei Wang

TL;DR

The paper tackles the computational bottleneck of spike sorting on high-channel-count neural probes by introducing E-Sort, an end-to-end neural-network sorter augmented with transfer learning and a GPU-friendly post-processing pipeline. The approach pre-trains temporal and spatial filters on large, neuron-rich recordings and fine-tunes them for a specific recording, significantly reducing the required labeled spikes and accelerating sorting. Key findings show up to $25.68\%$ accuracy improvement over training from scratch and a $2.25\times$ reduction in annotated spikes needed to reach Kilosort4-level performance, with the system sorting a 50-second recording in only $1.32$ seconds. Moreover, E-Sort generalizes across different probe geometries, noise levels, and drift conditions, and offers substantial runtime advantages over state-of-the-art sorters. This work suggests a practical path toward scalable, data-efficient spike sorting for modern neural probes.

Abstract

Decoding extracellular recordings is a crucial task in electrophysiology and brain-computer interfaces. Spike sorting, which distinguishes spikes and their putative neurons from extracellular recordings, becomes computationally demanding with the increasing number of channels in modern neural probes. To address the intensive workload and complex neuron interactions, we propose E-Sort, an end-to-end neural network-based spike sorter with transfer learning and parallelizable post-processing. Our framework reduces the required number of annotated spikes for training by 44% compared to training from scratch, achieving up to 25.68% higher accuracy. Additionally, our novel post-processing algorithm is compatible with deep learning frameworks, making E-Sort significantly faster than state-of-the-art spike sorters. On synthesized Neuropixels recordings, E-Sort achieves comparable accuracy with Kilosort4 while sorting 50 seconds of data in only 1.32 seconds. Our method demonstrates robustness across various probe geometries, noise levels, and drift conditions, offering a substantial improvement in both accuracy and runtime efficiency compared to existing spike sorters.

E-Sort: Empowering End-to-end Neural Network for Multi-channel Spike Sorting with Transfer Learning and Fast Post-processing

TL;DR

The paper tackles the computational bottleneck of spike sorting on high-channel-count neural probes by introducing E-Sort, an end-to-end neural-network sorter augmented with transfer learning and a GPU-friendly post-processing pipeline. The approach pre-trains temporal and spatial filters on large, neuron-rich recordings and fine-tunes them for a specific recording, significantly reducing the required labeled spikes and accelerating sorting. Key findings show up to accuracy improvement over training from scratch and a reduction in annotated spikes needed to reach Kilosort4-level performance, with the system sorting a 50-second recording in only seconds. Moreover, E-Sort generalizes across different probe geometries, noise levels, and drift conditions, and offers substantial runtime advantages over state-of-the-art sorters. This work suggests a practical path toward scalable, data-efficient spike sorting for modern neural probes.

Abstract

Decoding extracellular recordings is a crucial task in electrophysiology and brain-computer interfaces. Spike sorting, which distinguishes spikes and their putative neurons from extracellular recordings, becomes computationally demanding with the increasing number of channels in modern neural probes. To address the intensive workload and complex neuron interactions, we propose E-Sort, an end-to-end neural network-based spike sorter with transfer learning and parallelizable post-processing. Our framework reduces the required number of annotated spikes for training by 44% compared to training from scratch, achieving up to 25.68% higher accuracy. Additionally, our novel post-processing algorithm is compatible with deep learning frameworks, making E-Sort significantly faster than state-of-the-art spike sorters. On synthesized Neuropixels recordings, E-Sort achieves comparable accuracy with Kilosort4 while sorting 50 seconds of data in only 1.32 seconds. Our method demonstrates robustness across various probe geometries, noise levels, and drift conditions, offering a substantial improvement in both accuracy and runtime efficiency compared to existing spike sorters.
Paper Structure (12 sections, 1 equation, 3 figures, 1 table)

This paper contains 12 sections, 1 equation, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Dataset
  3. Methodology
  4. Dataset Preparation
  5. Transfer Learning
  6. Post-processing
  7. Experiments
  8. Experimental Setup
  9. Transfer Learning Results
  10. Generality of pre-trained Model
  11. Comparisons with State-of-the-art Spike Sorters
  12. Conclusion

Figures (3)

  • Figure 1: The proposed E-Sort spike sorting framework. (a) End-to-end NN pre-training & finetuning scheme. (b) Parallelizable post-processing for removing noises and redundant spikes.
  • Figure 2: Achieved accuracy versus number of spikes per neuron for finetuning pretrained models and training models from scratch.
  • Figure 3: Validations of recordings with different (a) probes, (b) noise levels, and (c) drifting types.