RippleNet

RippleNet detects sharp wave ripples (SPW-R) in hippocampal CA1 local field potential (LFP) recordings to identify events relevant to memory consolidation and decision-making.

Key Features:

• Recurrent Neural Network Architecture: Employs a deep recurrent neural network with Long Short-Term Memory (LSTM) layers to learn temporal dependencies in sequential LFP data for SPW-R identification.
• Self-improving AI Methodology: Learns features of SPW-R events directly from labeled datasets via supervised learning.
• Input and Output Specifications: Processes the low-frequency component of extracellularly recorded electric potentials (LFP) as input and outputs a time-varying probability trace for SPW-R occurrence that can be thresholded to determine event timings.
• Implementation Framework: Implemented using TensorFlow's Keras framework.

Scientific Applications:

• Memory consolidation research: Detection of SPW-R events in CA1 LFPs to support analysis of hippocampal replay and memory consolidation mechanisms.
• Decision-making and neural dynamics: Identification of SPW-Rs to study their role in decision-making and related hippocampal activity patterns.

Methodology:

The recurrent neural network with LSTM layers is trained via supervised learning on curated labeled datasets containing SPW-R events and learns features directly from raw LFP data without manual feature extraction.