HSST

HSST optimizes parameter selection for generic spike sorting algorithms to improve labeling and unit isolation of extracellularly recorded neural spike waveforms.

Key Features:

• Parameter optimization: Optimizes free parameter selection for generic spike sorting algorithms by evaluating multiple parameter sets.
• Neurophysiological priors: Incorporates known neurophysiological priors into parameter evaluation to favor physiologically plausible sorting outputs.
• Heuristic metrics: Uses heuristic neurophysiological metrics to evaluate and rank spike-sorting outputs.
• Unit isolation and signal discrimination: Selects parameter sets that enhance unit isolation and signal discrimination based on neurophysiological characteristics.
• Robustness across conditions: Demonstrates robustness across varying signal-to-noise ratios, numbers, and relative sizes of units per channel.
• Unsupervised operation: Performs parameter evaluation and selection in an unsupervised manner.
• Parallelization: Supports parallelized evaluation facilitating batch processing of datasets.
• Compatibility: Designed to work with generic spike sorting algorithms applied to extracellular microelectrode recordings.

Scientific Applications:

• Extracellular spike sorting: Optimizing spike sorting outputs for extracellular microelectrode recordings where signals from multiple neurons are captured simultaneously.
• Spike labeling accuracy: Improving assignment of detected spike waveforms to their originating neurons by selecting optimized sorter parameters.
• Parameter evaluation across conditions: Evaluating and selecting parameter sets across different signal-to-noise ratios and channel unit compositions.
• Large-scale dataset processing: Enabling batch processing of large neural recording datasets via parallelized parameter sweeps.

Methodology:

Evaluates ranges of input parameters for generic spike sorters using heuristic neurophysiological metrics and incorporated priors to identify physiologically plausible outputs; operates unsupervised and supports parallelized batch evaluation.