A State-Space Framework for Causal Detection of Hippocampal Ripple-Replay Events

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Hippocampal ripple-replay events are crucial for memory consolidation.
• Current methods for identifying these events require past and future data, hindering real-time analysis.
• Existing techniques fail to detect non-local representations without significant spectral changes in local field potentials (LFPs).

Purpose of the Study:

• To develop a temporally causal state-space model for real-time identification of hippocampal ripple-replay events.
• To enable the detection of non-local representations not accompanied by spectral changes in LFPs.
• To provide a framework for closed-loop experiments by enabling causal event detection.

Main Methods:

• Developed a novel state-space model integrating latent factors for neural oscillations, represented space, and coding property switches.
• The model simultaneously analyzes spiking activity from multiple units and LFP rhythmic content from multiple sources.
• Employed a temporally causal approach, allowing real-time state estimation using only past data, with options for future data refinement.

Main Results:

• The model successfully identified hippocampal ripple-replay events in simulated and real data.
• Demonstrated the ability to detect events causally, without reliance on future data.
• Showcased the model's capacity to identify concurrent changes in LFP rhythmic structure and place cell activity.

Conclusions:

• The proposed state-space model offers a causal and real-time method for identifying hippocampal ripple-replay events.
• This framework advances the study of memory consolidation and neural representations by overcoming limitations of previous methods.
• The model's causal nature opens new possibilities for real-time closed-loop neuroscience experiments.