Fuzzy assessment of sleep-disordered breathing during continuous positive airway pressure therapy

We propose a new method of quantifying sleep-disordered breathing (SDB) for the purpose of automating continuous positive airway pressure (CPAP) titration. Our algorithm, based on fuzzy logic, emulates the less-than-crisp kind of decision-making generally employed at the human level. Three input variables were first derived on a breath-by-breath basis from respiratory airflow measurements. These were: (1) the relative duration of inspiratory flow limitation in each breath; (2) the degree of hypopnea relative to the past 15 breaths; and (3) the intensity of snoring. Using these descriptors as inputs, our fuzzy inference algorithm produced a "severity index" (SI) quantifying the degree of SDB. Severity index was determined in CPAP titration procedures conducted on one normal snorer and 12 patients with moderate-to-severe obstructive sleep apnea. SI computed over the last 6 minutes of each CPAP level was compared against other more-conventional indices of SDB, such as total pulmonary resistance (RL), the number of apneas and hypopneas (NAH), and the number of arousals (NAr). In all but one of the subjects, the correlation coefficients for SI vs each of RL, NAH, and NAr were significantly different from zero, but not different from each other. The group correlation coefficients for SI vs RL, NAH, and NAr were 0.89, 0.86, and 0.87, respectively, demonstrating that SI accurately quantifies SDB. SI collapses multiple features of the airflow pattern into a single index and, therefore, may be useful as a "feedback" variable for the automatic control of CPAP therapy.