The electrocardiogram (ECG) is one of the most common ways to record, in an non-invasive manner, a patient’s cardiac activity. Once recorded the information can be pre-processed and subsequently analyzed to assess if the patient is suffering from any forms of cardiac abnormality which may require clinical intervention. In the current study we investigate ways in which more can be obtained from the ECG through analysis of the diagnostic properties of body surface potential maps (BSPM). A set of 192 lead BSPMs recorded from a mixture of 116 normal and abnormal subjects (59 normal vs 57 old myocardial infarction) were analyzed. For each patient, diagnostic features were obtained by calculating isointegral measurements from the QRS, STT, and entire QRST segments. These isointegrals provide a measure of the mean distribution of potential during ventricular depolarization, repolarisation, and a combination of both, respectively. For each isointegral type, 192 discrete measurements, and hence 192 features, were obtained; these correspond with the 192 leads recorded. Subsequent to this a signal-to-noise ratio-based feature ranking methodology was applied to select subsets of the best three, six and ten measurements (features) from the 192 available for each isointegral. These subsets of features were then applied to four different classifiers Naïve Bayes (NB), support vector machine (SVM), multi-layer perceptron (MLP) and random forest (RF) and in each application ten-fold cross validation was employed. It was found that when using the subsets of features obtained from the STT or QRST isointegrals, classification results in excess of 80% were attainable. This was in contrast to the results obtained using the QRS isointegral features where poorer performance (between 62.9% and 74.1%) was observed. The results from this study have illustrated that, for the studied dataset, the mean distribution of potentials during ventricular depolarization, and during ventricular repolarization and depolarization combined possessed greater diagnostic information. Overall it was concluded that this approach to BSPM analysis does provide a useful means for illustrating the usefulness of various features in diagnostic classification.