Interactive optimization of heuristic ECG classifiers

Diagnostic classification of the electrocardiogram (ECG) by computer is mostly carried out by means of heuristic classification algorithms, which mimic the cardiologist's way of reasoning. The authors developed a dedicated language, DTL (decision tree language) that allows cardiologists to express their classification knowledge in a way that is familiar to them. DTL can be processed by an interpreter which enables cardiologists to follow the classifier's operation and to interactively modify and test the algorithms. A translator is also available that can produce a fast run-time version of the program. The use of this environment to optimize the ECG computer program is described, and results of the improvement are given.