A Comprehensive Analysis of Machine Learning Techniques in Sleep Staging Systems Reveals Both Significant Benefits and Notable Challenges

Scientific research involving human subjects is conducted through various methods. The need for machine learning models to assist in designing and analyzing methods in the medical field is rapidly increasing. This paper presents different approaches to classification and regression techniques used in healthcare. While related review studies exist, this work seeks to offer new insights by comprehensively covering all aspects of classification and regression relevant to healthcare research. Traditionally, it involves manual classification by trained technicians who analyze the distinct characteristics of each sleep stage. This method is time-intensive, and the accuracy of staging can be influenced by subjective factors. Recently, there have been notable advancements in deep learning-based automatic sleep staging models. However, these models often do not account for age differences. This age-related shift makes older individuals more vulnerable to sleep disorders, making automatic sleep staging for this group more complex and deserving of increased focus. It focuses on selecting the most appropriate approach to address clinicians' challenges during implementation. Additionally, it outlines the progression from simple to complex models and discusses the challenges encountered when applying these approaches in practice.