t-SNE Visualization of Large-Scale Neural Recordings

Area of Science:

• Neuroscience
• Computational Neuroscience
• Big Data Analysis

Background:

• Electrophysiology is generating massive datasets, posing challenges for analyzing neuronal activity.
• Current spike sorting methods struggle with big data due to reliance on manual input.
• Accurate isolation of single neuron spikes is crucial for understanding brain function.

Purpose of the Study:

• To introduce t-SNE as a visualization tool for spike sorting in electrophysiology.
• To demonstrate the effectiveness of t-SNE in clustering and identifying single unit spikes.
• To provide a user-friendly graphical user interface (GUI) for spike sorting.

Main Methods:

• Applied t-distributed Stochastic Neighbor Embedding (t-SNE) to reduce dimensionality of extracellular spike data.
• Visualized high-dimensional spike features in a low-dimensional space to identify clusters.
• Tested the algorithm on labeled datasets from hybrid and paired juxtacellular/extracellular recordings.
• Developed a Python-based GUI for manual spike clustering and curation.

Main Results:

• t-SNE embeddings revealed obvious spike clusters, even from different feature spaces.
• These clusters accurately represent single units, validated on labeled datasets.
• The GUI facilitates precise manual delineation and curation of spike sorting results.

Conclusions:

• t-SNE is an effective tool for visualizing and sorting big data in electrophysiology.
• The developed GUI aids in accurate and efficient spike sorting, addressing limitations of current methods.
• Visualizations support the use of high-density probes and highlight the complexity of spike sorting across different recording conditions.