Related Experiment Video
Updated: Jul 1, 2025
10:35
Multi-layer Cortical Ca2+ Imaging in Freely Moving Mice with Prism Probes and Miniaturized Fluorescence Microscopy
Published on: June 13, 2017
31.1K
End-to-end deep learning approach to mouse behavior classification from cortex-wide calcium imaging
Takehiro Ajioka1, Nobuhiro Nakai1, Okito Yamashita2
1Department of Physiology and Cell Biology, Kobe University School of Medicine, Chuo, Kobe, Japan.
Plos Computational Biology
|March 13, 2024
Summary
Deep learning accurately decodes mouse behaviors from brain activity using a CNN-RNN model. This interpretable approach highlights key somatosensory cortex regions, advancing neural decoding in neuroscience.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Systems Neuroscience
Background:
- Deep learning models are increasingly used for neural decoding in neuroscience and clinical research.
- Interpretable deep learning models are essential for understanding how brain activity relates to behavior.
Purpose of the Study:
- To evaluate deep learning performance in classifying mouse behavioral states from calcium imaging data.
- To identify critical brain regions contributing to behavioral classification using an interpretable deep learning framework.
Main Methods:
- Utilized a convolutional neural network (CNN) combined with a recurrent neural network (RNN) for end-to-end behavioral state classification.
- Applied the CNN-RNN decoder to mesoscopic, cortex-wide calcium imaging data from mice.
- Analyzed the decoder's accuracy and robustness to individual differences on sub-second temporal scales.
Main Results:
- The CNN-RNN decoder achieved high accuracy and robustness in classifying mouse behavioral states.
- Identified significant contributions from forelimb and hindlimb areas of the somatosensory cortex to behavioral classification.
- Demonstrated the decoder's effectiveness on temporal scales of sub-seconds.
Conclusions:
- End-to-end deep learning approaches offer interpretable neural decoding with unbiased visualization of critical brain regions.
- The study provides a framework for linking specific brain activity patterns to observable behaviors.
- Findings advance the application of deep learning in systems neuroscience and clinical studies.
