Computer analysis of prolonged lower oesophageal sphincter pressure recordings

The aim of the study was to validate a recently developed computer program for the analysis of prolonged recordings of lower oesophageal sphincter pressure. Thirty 1-hour stretches were selected from sets of 24-h pressure signals recorded from the pharynx, oesophagus, lower oesophageal sphincter (LOS) and stomach in 10 ambulant patients with gastrooesophageal reflux disease. Three experienced investigators visually analysed end-expiratory LOS pressures and transient lower oesophageal sphincter relaxations (TLOSRs), using published criteria. A computer program was developed for calculation of an end-expiratory pressure curve and detection of TLOSRs using the same criteria. Although the results showed an maximum deviation from the mean of 11.1% and 14.8% for manually calculated LOS pressures and visually detected TLOSRs, respectively only 62.1% of the detected TLOSRs were detected by all three observers. LOS pressure as measured by the computer closely approximated the mean of the LOS pressures calculated by the three observers. Although the total number of TLOSRs was comparable to that assessed by visual analysis, the computer detected only 46% of the TLOSRs detected by each observer and 56.8% of the TLOSRs detected by all observers. It is concluded that automated calculation of end-expiratory LOS pressure is feasible and yields reliable results, whereas automated detection of TLOSRs could not be satisfactorily accomplished. Our study showed that improvement of computer algorithms for TLOSR detection is desirable. However, the previously described criteria for detection of TLOSRs are insufficiently precise; further refinement of these criteria will be necessary to reduce the large discrepancies between the outcome of detection of TLOSRs by computer and by humans, and to reduce the equally large discrepancies between the results of detection by different human observers.