PhysioEx: a new Python library for explainable sleep staging through deep learning.

Area of Science:

• Computational neuroscience
• Medical informatics
• Machine learning

Background:

• Sleep staging is vital for diagnosing sleep disorders.
• Current methods for sleep stage analysis can be complex and lack standardization.
• Deep learning (DL) and Explainable AI (XAI) offer potential for improved sleep analysis.

Purpose of the Study:

• Introduce PhysioEx, a Python library for sleep stage analysis.
• Provide a standardized, modular, and automated pipeline for sleep staging.
• Integrate DL and XAI for enhanced interpretability and clinical relevance.

Main Methods:

• Developed an extensible and modular API for sleep staging.
• Included pretrained models for various configurations (single-channel EEG, multi-channel EEG-EOG-EMG).
• Integrated XAI post-hoc methods to explain model decisions and align with clinical rules.

Main Results:

• PhysioEx benchmarks state-of-the-art sleep staging models.
• Enables fair comparison of models across different datasets.
• XAI techniques link DL decisions to human-understandable concepts, like American Academy of Sleep Medicine rules.

Conclusions:

• PhysioEx offers a standardized and accessible platform for sleep staging analysis.
• The library bridges the gap between machine learning models and clinical expertise.
• PhysioEx is a valuable, publicly available tool for sleep medicine researchers and practitioners.