The knowledge-based signal analysis for a heart sound information system

A heart sound information system (HSIS) has been developed, and a knowledge based signal analysis was employed in this study. We tried to combine knowledge engineering methods with traditional signal analysis methods for heart sounds to identify the heart sound components, classify the heart sounds, and make clinical diagnosis. Production rules were used to represent the knowledge in this system. In this system, we used backward-chaining when identifying heart sound components and used forward-chaining when making diagnosis. Metaknowledge was used to control the calling sequence of rules and to specify the problem solving strategies to enhance the inference effectiveness. To test the validity of knowledge based heart sound signal analysis, we selected 50 abnormal heart sound samples. The sources of these samples include those we collected from patients, those from existing heart sound tapes and those from other hospitals. Compared with the judgement of cardiologist, the coincidence rate was 86%.