Computational Neural Models for Body Surface ECG Data Analysis

Cardiovascular heart disease (CHD) is a chief public health priority worldwide. The 12-Lead Electrocardiogram (ECG) is a standard procedure in diagnosing CHDs such as Myocardial Infarction (MI). Nevertheless, due to sparse spatial sampling, it is limited in identifying cardiac abnormalities. Alternatively, in Body Surface Cardiac Mapping (BSCM) a higher number of ECGs are recorded. Hence, BSCM provides a more comprehensive picture of electrocardiographic information than is possible with the 12-lead ECG. This work has two main objectives. Firstly, to develop a classification framework for an accurate and early diagnosis of acute MI. This decision support system encompasses computational neural models with the input space based on BSCM. Secondly, since MI is localised on the torso surface, and due to the high number of electrocardiographic leads involved in BSCM, it is desirable to find an optimal reduced lead set for acute MI detection. By building an additional layer of knowledge between the cardiologist and clinical practice, this work not only enhances final MI classification performance but, allow the discovery of new electrocardiographic MI markers.