Model-based interpretation of cardiac beats by evolutionary algorithms

This work presents current progress of a new system for cardiac beat interpretation, combining model-based reasoning and evolutionary computing. As a difference from other model-based systems, the proposed approach directly integrates ECG signals with a cardiac model. Model-based reasoning is formalized as the minimization of an error function defined between the observation and the model's output. This paper presents a new adaptation process permitting the reproduction of a sequence of cardiac beats and the generation of beat interpretations by means of ladder diagrams. Results show the application of the proposed approach to two pathological rhythms from the MIT-BIH arrhythmia database. Although some limitations persist, preliminary results are encouraging and show the potentiality of this approach to exploit anatomic and physiological knowledge in order to explain underlying conduction disorders.