Evaluation of a root mean squared based ischemia detector on the long-term ST database with body position change cancellation

Abstract

In this work we revisit an ischemia detector based on the root mean square (RMS) series of the repolarization interval developed and validated using the European Society of cardiology ST-T database (ESCDB). This detector, developed within this database framework, gets sensitivity (S)/positive predictivity (+P) performance figures of 85%/86%. Our aim now is to re-evaluate the detector in the much richer long-term ST database where ST episodes of different origin are present, making a much more challenging scenario for the detector. Just a straight forward adaptation of the RMS detector reduces its performance figures, S/+P, to 70%/68%. This, apart from other reasons, is a consequence of the presence in the database of ST episodes generated by body position changes (BPC) which can be misinterpreted. A BPC detector incorporated to the previous detector noticeably improves the figures up to 75%/71%