Development and evaluation of a fully automated monophasic action potential analysis program

A fully automated program has been developed for analysing digitised cardiac monophasic action potentials (MAPs) of several animal species. Under diverse conditions, the program's performance is evaluated by comparison with high-resolution manual analysis of action potential duration, MAP detection, baseline and plateau determination. For each variable, the differences between both forms of analysis are categorised according to species (rabbit, dog or baboon), intervention (control, sematilide or LY190147), rhythm source (sinus or paced), and rates. They are then examined using ANOVA (null hypothesis: no difference between methods for any category). Intra-group differences are significant for baseline selection (p<0.05), and treatment effects are significant (p<0.05)for one species (baboon). However, the differences found are negligible when compared with expected measurement errors. Thus the program reliably detects and accurately analyses MAPS, independent of species, drug or pacing. The program's strength is its use of simple heuristic algorithms to identify and profile MAP signals while rejecting spurious transients. These algorithms focus on the procedural aspects of automatic MAP detection and validation, with minimum use of complex mathematics. The program's estimated throughput exceeds 8200 MAPs min(-1) on a 50 MHz i486 DX machine. With available data format conversion packages, signals from almost any data-acquisition source can be analysed.