Adaptive classification system for real-time detection of apnea and hypopnea events

This paper presents a real time portable apnea and hypopnea event detection system from the measurement of oronasal airflow signal only. The system uses a combined classification system for detection of events on the basis of personalized normal breathing pattern. Events are detected first, by identifying some abnormal breathing segments with the help of a binary classifier and then, the identified abnormal segments are further classified into any one of the two classes, i.e., apnea (A) and hypopnea (H). The second stage classification system is adaptive in nature, implemented to improve separation of apnea from hypopnea events. The proposed real time system was implemented in personal computer and was clinically validated by offline and online test, the event detection accuracy 93.4% and 91.8% was achieved on 8 different subjects in each case.