Automated analysis of paradoxical ribcage motion during sleep in infants

Identification of thoracoabdominal asynchrony (TAS) during breathing is currently detected by visual coding of records of ribcage (RC) and abdominal (AB) movements. There is thus a need to automate this process in order to save time and improve TAS detection accuracy. We studied 15 infants of 39-49 weeks postconceptional age. RC and AB signals were recorded continuously by inductance plethysmography for 4-24 hr immediately after herniorraphy. In our novel analysis approach, the records were divided into 10 sec epochs, and the equation RC = alphaAB + beta was fit to each epoch, using recursive linear regression with an exponential memory time constant of 1 and 2 sec. This yielded 10 sec signals for alpha corresponding to each epoch. The fraction of time that each alpha signal was positive was taken as a measure of synchrony between RC and AB for that epoch, while asynchrony was indicated by the fraction of time the signal was negative. We also assessed synchrony and asynchrony using a conventional measure known as thoracic delay (TD), which is based on the degree to which the peaks in RC and AB are coincident in time. Using TD as the basis of comparison, we found that our new recursive least squares method gave a positive predictive value of 99%. We conclude that our recursive least squares method is able to accurately identify portions of the RC and AB records that correspond to TAS, and we speculate that it may be useful in automating detection of TAS.