New models for knowledge-based interpretation and qualitative simulation of the electrocardiogram.

This thesis relates to the use of knowledge-based systems and fuzzy set theory in electrocardiogram (ECG) interpretation and qualitative ECG simulation. The aims of the research are (1) to design a knowledge-based system for ECG interpretation (KBS-EI) based on morphological features derived from ECGs, and (2) to explore a method for knowledge-based qualitative ECG simulation (KBS-QES) and its potential application in ECG interpretation. In KBS-EI, the architecture of a general blackboard system is used. KBS-EI contains several knowledge sources, each of which has its own inference engine and knowledge base. The knowledge needed for ECG interpretation is encoded in "nearly natural language" propositions which are organized into rules and frames. Fuzzy concepts in the knowledge are expressed in linguistic terms whose meanings are represented by corresponding membership functions. The way the encoded knowledge is used for ECG interpretation is based on modified "approximate analogical reasoning". Each interpretation of a set of 12-lead ECGs is assigned a similarity measure to indicate "confidence" in the interpretation. Currently, the system is being tested using ECGs obtained from medical monographs. Features of interest are extracted by manual measurement. The results so far show that the interpretations made by the system are highly consistent with those made by medical experts. Forty-five out of forty-nine cases were interpreted in a way consistent with the medical experts. KBS-QES is a system that performs qualitative ECG simulation using vectorial analysis. Simulated ECGs in each lead are generated by studying the projections of cardiac vectors in those leads. These projections are expressed in linguistic terms whose meanings are represented by corresponding membership functions. For a particular heart condition, 12-lead ECGs can therefore be simulated. In addition, such a simulation model can also be used to aid ECG interpretation. Case studies that were carried out show that under the various heart conditions considered, simulated ECGs were generated successfully. These studies also confirm that the system can aid ECG interpretation. KBS-QES is based on a multi-level architecture. This benefits the system by facilitating easy extension, modification and other future developments.