This repository contains code for the detection and classification of anesthesia-induced brain state transitions, including wakefulness, slow oscillations, and microarousals. We leverage a dual-model Convolutional Neural Network (CNN) and a self-supervised autoencoder-based multimodal clustering algorithm to achieve accurate brain state classification and transition detection based on in vivo LFP recordings from rats.
The pipeline processes the data through a series of steps, including preprocessing, state classification, and transition detection, using a combination of supervised and self-supervised learning techniques. It achieves accuracy rates of up to 96% for specific states and averages over 85% across all states, with 74% accuracy for detecting transitions. The methodology employs a leave-one-out strategy for model training, ensuring broad applicability across subjects.
For classification, check the example notebook located at example_notebook/general_notebook.ipynb. For transition detection, refer to the notebook at example_notebook/transition_notebook.ipynb. For further understanding and visualization of the process during the Autoencoders phase, please check and follow the comments in the example notebooks located at example_notebook/clusters_psd_notebook.ipynb and example_notebook/autoencoders_synthetic.ipynb.
Make sure to follow the instructions in the notebooks to properly preprocess your data, train the models, and perform the classification and transition detection tasks.
If you find this repository useful for your research, please consider citing our work:
For any questions or issues, feel free to raise an issue on this GitHub repository, and we will do our best!