Multi-component based cross correlation beat detection in electrocardiogram analysis

A new beat detection approach is proposed based on the fundamentals of cross correlation and compared with two benchmarking approaches of non-syntactic and cross correlation beat detection. The new approach can be considered to be a multi-component based variant of traditional cross correlation where each of the individual inter-wave components are sought in isolation as opposed to being sought in one complete process. Each of three techniques were compared based on their performance in detecting the P wave, QRS complex and T wave in addition to onset and offset markers for 3000 cardiac cycles.Results indicated that the approach of multi-component based cross correlation exceeded the performance of the two benchmarking techniques by firstly correctly detecting more cardiac cycles and secondly provided the most accurate marker insertion in 7 out of the 8 categories tested.The main benefit of the multi-component based cross correlation algorithm is seen to be firstly its ability to successfully detect cardiac cycles and secondly the accurate insertion of the beat markers based on pre-defined values as opposed to performing individual gradient searches for wave onsets and offsets following fiducial point location.Computerised classification of the electrocardiogram (ECG) is a complex and multi staged process. The overall goal is to determine if the patient is 'normal' and may remain untreated, or whether the patient exhibits any cardiac abnormalities requiring treatment. The classification of the ECG by computerised techniques has been an active area of research for more than 4 decades. A plethora of algorithmic techniques have been applied and developed [1,2] all with the common goal of enhancing the classification accuracy and becoming as reliable and successful as expert cardiologists. The process can be divided into a number of sequential stages. Pre-processing stages of Beat Detection and Feature Extraction/Selection provide suitable information from the recorded