A new fuzzy based diagnosing system for instantaneous processing 12 lead ECG signal

The Electrocardiogram (ECG) signal reflects the performance of the human heart as an electrical signal. It consists of three main waves (P, QRS complex, and T), and is recorded by an ECG machine in the form of 12 leads which include valuable information about the functional activities of the huma...

Full description

Saved in:
Bibliographic Details
Main Author: Sameer Kleban, Salih
Format: Thesis
Language:English
Subjects:
Online Access:http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/59423/1/Page%201-24.pdf
http://dspace.unimap.edu.my:80/xmlui/bitstream/123456789/59423/2/Full%20text.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The Electrocardiogram (ECG) signal reflects the performance of the human heart as an electrical signal. It consists of three main waves (P, QRS complex, and T), and is recorded by an ECG machine in the form of 12 leads which include valuable information about the functional activities of the human heart and cardiovascular system. It is annotated manually by a cardiologist to diagnose cardiac disease, but for a long time ECG recordings were performed to get an effective measure of heart rate variability. The generated ECG data is huge and the probability of wrong analysis or misreading by manual annotation is increased. Therefore, many computerized based techniques have been proposed in literature for analyzing and detecting ECG waves, and at a lower rate for diagnosing cardiac diseases. In this thesis, a new robust intelligent system has been proposed to perform an accurate diagnosis of a high risk cardiac disease named left ventricular hypertrophy (LVH). Four approaches are developed within the proposed ECG system to improve the performance of processing the ECG signal with respect to the existing methods and to discover new system for diagnosing cardiac disease based on the computerized intelligence technique. The first proposed approach is a digital recovery system which addresses the limitation of digital 12 lead ECG data by reconstructing it from the colour scanned image of the ECG printed chart. This approach is implemented by four image processing steps and captures raw ECG data with respect to the baseline which is detected by the same approach. Furthermore, it is reliable for different ECG morphologies and printout charts. The reconstructed data is evaluated qualitatively and quantitatively using some predefined standard features. The analytic results demonstrate the consistency and robustness of this approach to generate 12 lead ECG data with high precision (98%). The second and third approaches are proposed to detect ECG waves and then delineate all time characteristics of these waves. In contrast to the existing methods, both approaches are based on straightforward algorithms that perform instantaneous processing for the ECG signal. As a result, detection operation is executed in a high speed which reaches (4.5s per 650,000 beats) for QRS complex and (2.7s per 225,000 beats) for P&T waves. The based technique in both detection approaches has the advantage of rising falling edge mutation as a base rule for delineating subject. This technique reduces undetected beats and provides accurate detection results exceeding ones in up to date existing methods. The fourth proposed approach is a diagnostic system for LVH cardiac disease based on proposed diagnostic criterion. In contrast to the conventional LVH diagnostic criteria, the decision in the proposed criterion is computed by three logical expressions; two of which are determined by a combination of classic criteria, whereas the third is obtained by eight ECG voltages and takes two different levels for each gender. These expressions are represented by the membership functions in the proposed design of the fuzzy inference system. The proposed diagnosing system is validated by fifty ECG records, in which the validation results score were perfect (100%) in terms of sensitivity, specificity, and accuracy, while the best diagnosing accuracy achieved by traditional diagnostic criteria does not exceed 90%.