Parametric Estimation of Respiratory Signals for ML-Based Early Detection of Sleep Apnea

Sleep apnea is a serious and widespread condition in which breathing is repeatedly interrupted during sleep. Despite its high prevalence, it remains significantly underdiagnosed due to the reliance on expensive and timeconsuming clinical procedures like overnight polysomnography. To clear this gap, we developed a machine learning-based algorithm for the early detection of sleep apnea using simplified respiratory signal analysis. Our strategy is in the modeling of both nasal airflow and thoracic effort signals using parametric techniques with the extraction of pertinent features such as changes in breath depth, frequency, and signal irregularities. The features are then used for training and testing of several machine learning models such as Random Forest, Support Vector Machine, and Long Short-Term Memory (LSTM) networks. Among them, the LSTM model showed the highest performance, recognizing apnea events with both high sensitivity and reliability. The system was additionally tested with external datasets and checked for robustness against noisy inputs. The results show that our method is accurate, interpretable, and computationally efficient for real-time use. The study offers a practical and scalable solution for noninvasive, early screening of sleep apnea outside standard clinical practice.