Detection of Sleep Apnea using Cat Swarm Optimization with TabNet versus CNN Model

Sleep apnea is a sleep disease that is often an undetected cause of death. It is generally diagnosed with polysomnography, which can be costly and uncomfortable. To solve this issue, we suggest a website for detecting sleep apnea that uses user-provided data such as the Berlin questionnaire, Epworth Sleepiness Scale (ESS), body measurements, Body Mass Index (BMI), age, and gender. The purpose of this work is to categorize sleep apnea based on the Apnea-Hypopnea Index (AHI) as normal, mild, moderate, or severe. This classification is carried out using a deep learning model, TabNet which suits the best to process the text input. It could be useful for people who feel they have sleep apnea, allowing them to take the essential steps toward a formal diagnosis and therapy. However, this aids in early diagnosis by indicating the stage of the disease, and users are recommended to check with healthcare specialists for a definitive diagnosis. For feature selection, nature-inspired Cat Swarm optimization is applied. The TabNet is trained on a dataset that includes these characteristics as well as the AHI, a common metric for determining the severity of sleep apnea. The algorithm uses these inputs to forecast the likelihood of having sleep apnea: normal, mild, moderate, or severe. The developed TabNet Model is compared with the Convolution Neural Network (CNN) and the results are showcased. To summarize, this approach is a revolutionary technique for sleep apnea detection using Cat Swarm Optimization. It promises to give a more accessible and cost-effective solution for people who feel they have this illness by utilizing user-generated data and deep learning algorithms.