An experience-based multi-lead decision model for electrocardiogram wave boundary detection

An experience-based multi-lead (12 standard leads) decision model was presented for locating the ECG wave boundary. After getting 12 single-lead ECG boundary results from any single-lead detector (used threshold based method), the model first applied a data selecting and alignment algorithm to filter invalid records in each beat. Then valid data were assigned to different weights in each lead for calculating a final location according to the rule approved by physicians. This method has been assessed in our ECG database. The total accuracy was 92.4% and mean deviation was 4.2ms between the standard marking and our algorithm's marking. The software with this method has been applied in remote medical center, in which a good feedback was given through the test of a large amount of actual data..