Sleep apnea detection using flow spectral analysis and fuzzy logic

Diagnosis of sleep apnea is currently performed by a full night polysomnography study at sleep laboratories. The majority of apnea patients are then treated by constant positive airway pressure (CPAP) device. A new algorithm to detect and classify apnea that could be easily incorporated into CPAP was developed. The study used data from ten subjects (6 males, 4 females, age mean (SD): 52 (13), BMI mean (SD): 34 (8)) who were previously diagnosed with sleep apnea. The CPAP flow data was used for the algorithm development. Spectral analysis of data was performed and area, weighted mean frequency and amplitude were determined. These were fed as inputs to a fuzzy logic program, which was used to detect normal, obstructive sleep apnea (OSA) and central sleep apnea (CSA). The data analyzed included 591 normal breaths and 306 apnea events (165 obstructive and 140 central). Of these 299 normal breaths and 148 apnea events (77 obstructive and 71 central) were used for algorithm development and the remainder for testing. The algorithm had correct detection rates of 99.6% for normal breaths, 69.6% for CSA and 71.9% for OSA. The results suggest that the algorithm could be successfully used to determine if a subject had apnea or not. It has limited success in differentiating apnea types.