Artificial Intelligence Models for the Automation of Standard Diagnostics in Sleep Medicine—A Systematic Review

Abstract

Sleep disorders, prevalent in the general population, present significant health challenges. The current diagnostic approach, based on a manual analysis of overnight polysomnograms (PSGs), is costly and time-consuming. Artificial intelligence has emerged as a promising tool in this context, offering a more accessible and personalized approach to diagnosis, particularly beneficial for under-served populations. This is a systematic review of AI-based models for sleep disorder diagnostics that were trained, validated, and tested on diverse clinical datasets. An extensive search of PubMed and IEEE databases yielded 2114 articles, but only 18 met our stringent selection criteria, underscoring the scarcity of thoroughly validated AI models in sleep medicine. The findings emphasize the necessity of a rigorous validation of AI models on multimodal clinical data, a step crucial for their integration into clinical practice. This would be in line with the American Academy of Sleep Medicine’s support of AI research.

Artificial Intelligence in Sleep Medicine: The Dawn of a New Era

“…Nevertheless, thoroughly validating AI models using diverse clinical data as a vital step for their adoption in clinical settings is still needed. 5 …”

“…31 The selection of appropriate training data is vital for developing accurate and generalizable AI models in sleep medicine research, as the size, quality, and representativeness of the dataset can significantly affect model performance and the likelihood of overfitting or underfitting, which are issues related to how well the model predicts new, unseen data compared to how it performs on the training data. 5 To mitigate these risks, researchers should strive to use large, diverse, and representative datasets for training AI models, employ random sampling methods to guarantee that the data accurately reflects the target population, and apply iterative model training along with independent validation to evaluate the stability and generalizability of the models developed. 32 Initiatives such as the National Sleep Research Resource (NSRR) and the UK Biobank have made large sleep datasets publicly available, enabling researchers worldwide to leverage this valuable data for analysis.…”

“…Nevertheless, thoroughly validating AI models using diverse clinical data as a vital step for their adoption in clinical settings is still needed. 5 Building on these advancements, the FDA has cleared a few auto-scoring software systems for use, 6 and the American Academy of Sleep Medicine (AASM) has initiated a two-year pilot program to certify autoscoring software,…”