IMPROVED MODEL FOR THE DETECTION OF MYOCARDIAL INFARCTION FROM MULTILEAD ECG USING QRS POINT SCORE AS AN ADDITIONAL FEATURE

Complexities of interpretation of various ECG findings in patients with myocardial infarction are well known. This study is an attempt to find out the utility of point scoring system in diagnosing myocardial infarction. The present study was done as an analysis of the data available in the database PTB from the public domain "Physionet" where 12 lead simultaneous signals of Normal patients and Myocardial Infarction are available. Multi-lead ECGs acquired simultaneously improves the accuracy in the diagnosis of heart diseases. The signals were analyzed for each of the 34 normal patients and 33 patients who have been diagnosed to have myocardial infarction. Point score as a feature and Naïve Bayes classifier were used to assess the ECGs. The point scores and Naïve Bayes classifier found the maximum diagnostic accuracy in the lead V6 where the area under curve is 0.968 and 95.65% individuals were correctly classified. Kappa score for all the leads when both the point score and Naïve Bayes classifier was used ranged between 0.78 and 0.96 with 93% sensitivity but with the exclusion of the point scores, the same ranged between 0.61 and 0.87. We found the combination of both point scores and Naïve Bayes classification to be good predictive utility in diagnosing myocardial infarction.