rolling correlation.pdf

IOP PUBLISHING PHYSIOLOGICAL MEASUREMENT Physiol. Meas. 30 (2009) 1241–1250 doi:10.1088/0967-3334/30/11/008 Automated identification of peristaltic pressure waves in oesophageal manometry investigations using the rolling correlation technique S Perring and E Jones Medical Physics Department, Poole Hospital NHS Trust, Longfleet Rd, Poole BH15 2JB, UK E-mail: Steve.perring@poole.nhs.uk Received 9 July 2009 accepted for publication 7 September 2009 Published 8 October 2009 Online at stacks.iop.org/PM/30/1241 Abstract We have implemented the technique of rolling correlation coefficient as proposed by Buttfield and Bolton (2005 Real time measurement of RR intervals using a digital signal processor J. Med. Eng. Technol. 29 8–13) for ECG R-wave detection in the detection and timing of oesophageal peristalsis. 43 sequential patients attending for oesophageal manometry were retrospectively reviewed. Two expert reviewers visually assessed each swallow for normality of peristaltic amplitude and propagation speed. Automatic assessment was performed using rolling correlation, maximum amplitude, threshold and maximum gradient techniques of identifying onset of peristalsis. Rolling correlation was comparable with the maximum amplitude technique at identifying peristaltic pressure waves visually identified as present. Rolling correlation was most effective at correctly identifying propagation velocity as normal (698 out of 845 normally propagating waves) and highest correlation with expert visual assessment of percentage abnormal propagation for each patient (R value 0.918). In a sub-group of 11 studies assessed as displaying normal motility, rolling correlation gave lowest variation of propagation speed and highest consistency with visual assessment. The rolling correlation technique is effective and accurate at identifying oesophageal peristalsis and characterizing peristaltic propagation in manometric studies even in the presence of abnormally weak peristalsis and other confound