A Comparative Study of CNN and SVM for Sleep Apnea Classification: Evaluating Performance and Efficiency

Inadequate oxygen levels can happen when a person with sleep apnea has periods of paused breathing and then resumes breathing during the night. If the signal is not received, the individual's airway will close and they will cease to breathe. These interruptions can occur occasionally but frequently during sleep. When they start breathing again, they could snort, inhale deeply, or wake up feeling as though they are choking, gasping, or suffocating. This study introduces a classification model to identify different types of sleep apnea using polysomnographic data, with a focus on Normal (No Apnea), Obstructive Sleep Apnea (OSA), Central Sleep Apnea (CSA), Mixed Sleep Apnea (MSA), and Hypopnea (Reduced Breathing). The model showed excellent accuracy levels, with precision and recall measurements demonstrating strong performance, especially in Normal and Hypopnea categorizations. Although effective, the model encountered difficulties in correctly differentiating between CSA and MSA, as shown by higher rates of false negatives in these groups. the model achieves an impressive accuracy of 99% for Normal and OSA, and 97% for CSA and MSA, demonstrating its efficacy in classifying sleep apnea types and underscoring its potential for clinical application in diagnosing sleep disorders.