Irregularity and Variability Analysis of Airflow Recordings to Facilitate the Diagnosis of Paediatric Sleep Apnoea-Hypopnoea Syndrome

The aim of this paper is to evaluate the evolution of irregularity and variability of airflow (AF) signals as sleep apnoea-hypopnoea syndrome (SAHS) severity increases in children. We analyzed 501 AF recordings from children 6.2 ± 3.4 years old. The respiratory rate variability (RRV) signal, which is obtained from AF, was also estimated. The proposed methodology consisted of three phases: (i) extraction of spectral entropy (SE1), quadratic spectral entropy (SE2), cubic spectral entropy (SE3), and central tendency measure (CTM) to quantify irregularity and variability of AF and RRV; (ii) feature selection with forward stepwise logistic regression (FSLR), and (iii) classification of subjects using logistic regression (LR). SE1, SE2, SE3, and CTM were used to conduct exploratory analyses that showed increasing irregularity and decreasing variability in AF, and increasing variability in RRV as apnoea-hypopnoea index (AHI) was higher. These tendencies were clearer in children with a higher severity degree (from AHI ≥ 5 events/hour). Binary LR models achieved 60%, 76%, and 80% accuracy for the AHI cutoff points 1, 5, and 10 e/h, respectively. These results suggest that irregularity and variability measures are able to characterize paediatric SAHS in AF recordings. Hence, the use of these approaches could be helpful in automatically detecting SAHS in children.