SHYBRID

SHYBRID generates hybrid ground-truth spiking data by relocating spikes from recorded single units to new recording-probe locations to create virtual units with known spike times for evaluation of spike sorting algorithms on extracellular neural recordings.

Key Features:

• Hybrid ground-truth data generation: Relocates spikes from a single unit to different locations on the recording probe to produce virtual units with known spike times.
• Relocation-based virtual unit creation: Produces controlled, synthetic units within real extracellular recordings while preserving original spike waveforms and timing.
• Algorithm comparison and parameter optimization: Provides identical controlled datasets to compare multiple spike sorting algorithms and to fine-tune their parameters.
• Controlled perturbations for robustness testing: Enables manipulation of noise levels and signal overlap to evaluate algorithm performance under varied recording challenges.
• Data-driven modeling and implementation: Implements a data-driven modeling paradigm in Python for generation and manipulation of hybrid ground-truth data.

Scientific Applications:

• Spike sorting evaluation: Benchmarking and validation of spike sorting algorithms using datasets with known spike times.
• Algorithm optimization: Systematic tuning of spike sorting parameters to improve detection and classification accuracy for specific recording settings.
• Methodological studies: Investigation of algorithm failure modes and sensitivity to noise, signal overlap, and probe geometry.
• Ground-truth dataset generation for neuroscience: Creation of validated spiking datasets to support development and comparison of computational methods for neural data analysis.

Methodology:

Relocation of spikes from single units to different recording-probe locations to create virtual units with known spike times, implemented within a data-driven modeling paradigm in Python.