November 8th, 2024
We present a software package with a graphic user interface for researchers without coding experience to score sleep stages in mice with a simple download and operation.
We developed an easy-to-use GUI software, IntelliSleepScorer, to automatically score sleep stages in mice. We aim to address the challenge of labor-intensive manual scoring and improve accessibility for researchers without much coding experience. IntelliSleepScorer achieves 95.2%accuracy in automatically scoring sleep stages in mice, and proved generalizability on two other publicly available independent datasets.
The GUI includes a manual correction feature, allowing users to modify the scored sleep stage for any Epoch. We also provide a SHAP Value Visualizations to explain the model scoring decisions. Experienced users can also fine tune our pre-tree model with their own data as all necessary files and extraction scripts are available on GitHub.
Our research focuses on translating emerging genetics into biology and enabling next-generation therapeutics to treat psychiatric illnesses. We have been working on genes whose dysfunction has been implicated on these illnesses using molecular, cellular, and electrophysiological approaches, both in vitro and in animals. To begin, convert the file format into EDF and organize them into LightGBM 2EEG, and LightGBM 1EEG models.
Then download the IntelliSleepScorer software, as per your operating system. For Windows, download the executable file through the Pan Group Research page. For macOS or Linux, download the source code from the GitHub Repository.
Download two example datasets saved as EDF files through the GitHub Repository. Next, download the models. zip file, unzip it, and copy the extracted models folder inside the IntelliSleepScorer Repository.
To launch the IntelliSleepScorer software in Windows, double click the IntelliSleepScorer. exe in the root folder. For macOS or Linux, open a terminal emulator.
Change the directory to the root folder and enter the command. In the software interface, click, Select EDF or EDF+files to choose the files for scoring. Select the desired Epoch Length using the dropdown menu.
Choose from four, 10, or 20 seconds for Sleep Stage Scoring. Then select the model for sleep scoring, based on the file structure. Choose LightGBM-2EEG for data files with two EEG channels and one EMG channel.
Check the Run or Plot SHAP checkbox to include the SHAP Calculation before running the prediction. Next, click, Score All Files, to score the selected EDF or EDF+files. The model will automatically score the files and calculate Global and Epoch SHAP values to interpret the scoring decisions.
After completing the sleep scoring process, click, Visualize the Selected File, to display the EEG and EMG signals and a Hypnogram aligned with the signals. To change the number of Epochs displayed, click the dropdown menu next to select number of Epochs to display and select the desired value. Use the navigation buttons to move forward and backward through different Epoch data.
To view Epoch Level SHAP Values, right click on a specific Epoch to plot the SHAP data. Use the pink transparent bar on the left side of the Hypnogram to indicate the current location of the Epoch. To switch to another Epoch, left click on the Hypnogram or click, Go to Epoch, and enter the Epoch number.
If SHAP Calculation is enabled right click the selected Epoch to generate its Epoch SHAP Plot. To manually correct a predicted sleep stage, left click on an EPOCH in the Sleep Stage Prediction Plot to select it. The model's predicted stage for the selected Epoch will appear on the right.
Click on the Widget and select a New Stage from the Wake, NREM and REM options in the dropdown menu. Finally, to reopen a previously saved scored file, ensure the Epoch Length and selected model match the settings used during the initial processing.
This study presents IntelliSleepScorer, an easy-to-use graphical user interface (GUI) software designed for researchers to automatically score sleep stages in mice without requiring coding skills. The software demonstrates a high accuracy of 95.2% in scoring and allows for manual corrections and SHAP value visualizations to explain scoring decisions.
Automated, high-accuracy sleep stage scoring in rodent models is critical for translational neuroscience and neuropsychiatric drug discovery. IntelliSleepScorer enables scalable, reproducible quantification of sleep architecture, supporting mechanistic de-risking and target validation in preclinical research. Its accessible interface and explainable outputs facilitate cross-functional adoption and portfolio-wide data standardization.
IntelliSleepScorer integrates into the preclinical discovery continuum, from early target validation through lead identification and translational research in neurobiology.