A Systematic Review and Meta-Analysis on Sleep Stage Classification and Sleep Disorder Detection Using Artificial Intelligence

Overview

• This paper provides a comprehensive review and meta-analysis of recent research on the use of artificial intelligence (AI) in sleep studies, including the classification of sleep stages and the detection of sleep disorders.
• The study examines various AI models, data sources, and performance metrics used in these studies, highlighting the contributions of AI to sleep research.
• The findings aim to help physicians and researchers understand the current state of AI-based approaches to sleep analysis and their potential for improving clinical assessments.

Plain English Explanation

Sleep is crucial for our physical and mental health, and understanding individual sleep patterns and disorders can provide valuable insights into overall well-being. This review paper summarizes the latest research on using artificial intelligence to analyze sleep data and classify sleep stages and disorders.

Traditionally, sleep studies have relied on polysomnography (a detailed recording of various body functions during sleep) and self-reported measures. However, the evaluation process has become more convenient with the advancements in AI, which can help classify sleep stages and detect sleep disorders more accurately and efficiently.

This paper reviewed 183 articles on AI-based sleep studies, focusing on 80 of them published between 2016 and 2023. The researchers found that brain waves were the most commonly used body parameters for these studies, and the convolutional neural network was the most widely used AI model, accounting for 27% of the studies.

Other AI models, such as long short-term memory, support vector machine, random forest, and recurrent neural network, were also employed in the reviewed studies. The researchers assessed the performance of these models using various metrics, including accuracy, F1 score, kappa, sensitivity, and specificity.

The findings of this review can help physicians and researchers understand how AI is being used to study sleep patterns and disorders, and potentially improve the efficiency and accuracy of clinical assessments. For example, contactless polysomnography using radio waves or AI-based detection of Alzheimer's disease from sleep patterns are some of the innovative applications highlighted in the review.

Technical Explanation

The paper systematically reviewed 183 articles related to the use of AI in sleep studies, with a focus on 80 records published between 2016 and 2023. The researchers examined the different approaches, data sources, AI models, and performance metrics used in these studies.

The most commonly used body parameter for sleep staging and disorder detection was brain waves, which were employed in the majority of the reviewed studies. The convolutional neural network (CNN) was the most widely used AI model, accounting for 27% of the studies. Other models, such as long short-term memory (LSTM), support vector machine (SVM), random forest (RF), and recurrent neural network (RNN), were also utilized in the reviewed research.

In terms of performance metrics, accuracy was the most widely used, with a maximum of 83.75% reported. Other metrics, such as the F1 score (45%), kappa (36.25%), sensitivity (31.25%), and specificity (30%), were also employed to assess the performance of the AI models.

The review highlights the growing use of AI in sleep research, including innovative applications like clustering and data augmentation techniques to improve sleep stage classification accuracy and novel hybrid self-supervised learning frameworks for sleep staging. These advancements have the potential to enhance the efficiency and accuracy of clinical assessments, benefiting both physicians and patients.

Critical Analysis

The review provides a comprehensive overview of the current state of AI-based sleep studies, highlighting the various approaches, models, and performance metrics used in this field. However, the paper does not delve deeply into the limitations and potential issues associated with these AI-based techniques.

For example, the review does not address the challenges of data quality and diversity in sleep studies, which can significantly impact the performance of AI models. Additionally, the paper does not discuss the ethical considerations and potential biases that may arise from the use of AI in clinical decision-making, which is an important aspect to consider.

Furthermore, while the review covers a wide range of studies, it does not provide a critical analysis of the strengths and weaknesses of the different AI models and techniques employed. A more in-depth discussion of the trade-offs and suitability of various approaches for different sleep-related tasks would have been valuable.

Clustering and data augmentation techniques or novel hybrid self-supervised learning frameworks mentioned in the review could have been further explored to understand their limitations and potential areas for improvement.

Overall, while the review provides a useful summary of the current state of AI in sleep research, a more critical and analytical approach would have strengthened the paper's contribution to the field.

Conclusion

This comprehensive review and meta-analysis of recent literature on the use of AI in sleep studies highlights the growing importance of this technology in advancing our understanding of sleep patterns and disorders. The findings demonstrate the versatility of AI models, such as CNNs, LSTMs, SVMs, and RNNs, in classifying sleep stages and detecting sleep-related issues.

The review's insights can help physicians and researchers stay updated on the latest AI-based approaches and their potential for improving clinical assessments. As the field continues to evolve, future research should address the limitations and ethical considerations surrounding the use of AI in sleep analysis, ensuring that these advancements positively impact patient care and outcomes.