A Method for Quantifying Similarity Between Population Spiking Activities

• Variance-weighted neural similarity analytics delivering noise-corrected, reproducible population spike comparisons.

Abstract

As neural recording systems have expanded in channel density and data complexity, reliably comparing population-level spike activity across trials, sessions, and subjects has remained a significant challenge. Traditional PCA- and CCA-based methods have frequently overstated similarity due to smoothing artifacts, unequal variance contributions, and stochastic noise. Researchers at USC have developed ReBaCCA-ss to address these limitations by integrating variance-balanced alignment, surrogate-based baseline correction, and automated smoothing optimization into a unified analytical framework. This approach has generated bounded, interpretable similarity metrics that have strengthened analytical reliability in large-scale neural datasets.

Benefit

• Reduces similarity inflation caused by smoothing artifacts and random spike correlations.

• Provides variance-weighted similarity metrics that improve benchmarking and model validation.

• Integrates into scalable neural analytics pipelines for high-channel-count recording systems.

Market Application

Researchers at USC have designed ReBaCCA-ss to support commercial applications in brain-computer interfaces, adaptive neuromodulation systems, neural diagnostics, and AI-driven neural analytics. The framework has enabled longitudinal neural monitoring, cross-session and cross-subject alignment, and validation of neural decoding models. The technology supports performance benchmarking, regulatory documentation, translational validation, and product differentiation across invasive and non-invasive neural interface markets where reliable population-level similarity metrics are required for commercialization.

Publications

Zhang, Xiang, et al. “ReBaCCA-ss: Relevance-Balanced Continuum Correlation Analysis with Smoothing and Surrogating for Quantifying Similarity Between Population Spiking Activities.” Neural Computation, manuscript submitted, 2025.

Zhang, Xiang, and Dong Song. “ReBaCCA-ss: Relevance-Balanced Continuum Correlation Analysis with Smoothing and Surrogating.” arXiv, 2025. https://arxiv.org/abs/2505.13748

Other

• Validated software prototype tested on simulated datasets and in vivo hippocampal multi-electrode recordings from rats performing a memory-guided Delayed Nonmatch-to-Sample task.

• Demonstrated measurable performance improvements over conventional CCA-based approaches in relevant research environments.

• Available for licensing (exclusive or non-exclusive).