Integrating dynamic Bayesian networks and constraint-based fuzzy models for myocardial infarction classification with 12-lead ECGS

This paper presents a novel combination of the dynamic Bayesian networks (DBNs) and constraint-based fuzzy models for myocardial infarction classification with 12-lead ECGs. Data of lead-V1, V2, V3, V4 were selected. Then, DBNs were used for finding the likelihood value which was treated as statistical feature data of each heartbeat´s ECG complex, and constraint-based fuzzy models were used to extract knowledge from the trained DBNs. The fuzzy model developed from this approach is tested on 905 samples of heartbeats from clinical data, including 470 data with myocardial infarction and 435 data from healthy individuals. The sensitivity of the classifier achieved 86.27% and prediction accuracy achieved 78.32%.