RESP-24: a computer program for the investigation of 24-h breathing abnormalities in heart failure patients

In this paper, we describe a computer program (RESP-24) specifically devised to assess the prevalence and characteristics of breathing disorders in ambulant chronic heart failure patients during the overall 24 h period. The system works on a single channel respiratory signal (RS) recorded through a Holter-like portable device. In the pre-processing stage RESP-24 removes noise, baseline drift and motion artefacts from the RS using a non-linear filter, enhances respiratory frequency components through high-pass filtering and derives an instantaneous tidal volume (ITV) signal. The core processing is devoted to the identification and classification of the breathing pattern into periodic breathing (PB), normal breathing or non-classifiable breathing using a 60 s segmentation, and to the identification and estimation of apnea and hypopnea events. Sustained episodes of PB are detected by cross analysis of both the spectral content and time behavior of the ITV signal. User-friendly interactive facilities allow all the results of the automatic analysis procedure to be edited. The final report provides a set of standard and non-standard parameters quantifying breathing abnormalities during the 24 h period, the night-time and the day-time, including the apnea/hypopnea index, the apnea index, the total time spent in apnea or in hypopnea and the prevalence of non-apneic and apneic PB. The accuracy of these measurements was appraised on a data set of 14 recordings, by comparing them with those provided by a trained analyst. The mean and standard deviation of the error of the automatic procedure were below respectively 6 and 8% of the reference value for all parameters considered and the mean total classification accuracy was 92%. In most cases, the individual error was <12%. We conclude that measurements provided automatically by the RESP-24 software are suitable for screening purposes and clinical trials, although a preventive check of signal quality should be recommended.