An integrated index to detect the extent of myocardial infarction using echocardiographic texture analysis
Date of Issue2016
School of Mechanical and Aerospace Engineering
Expansion of infarcted zone after Myocardial Infarction (MI) has serious short and long-term consequences and contributes to increased mortality. Early detection and quantitative assessment of MI will help to prevent further damage of the cardiac muscles. Though two-dimensional (2D) echocardiogram helps to detect MI at early stage, the reading of ultrasound images is subjective due to inter-observer variability and may lead to inconclusive findings which may increase the patient anxiety and cost of treatment. Automated detection of MI and quantification of the extent of MI will enhance the capabilities of diagnosis system. However, little work has been done in the area of computer-based texture analysis for automated detection of extent of MI in general, and existing literature is found to be largely confined only to MI detection. In this thesis, various algorithms have been proposed to identify infarcted myocardium and extent of infarcted myocardium that are categorised into (a) normal myocardium, (b) moderately infarcted myocardium, and (c) severely infarcted myocardium using 2D echocardiogram texture analysis. The automated identification of normal and infarcted myocardium algorithm is developed first, using Discrete Wavelet Transform (DWT), second order statistics calculated from Gray- Level Co-ccurrence Matrix (GLCM) and Higher-Order Spectra (HOS) texture descriptors. The proposed system is validated using 400 MI and 400 normal ultrasound images, obtained from 80 patients with MI and 80 normal subjects. Among the three techniques presented, DWT yielded the highest classification accuracy. Another efficient multi-resolution decomposition algorithm is proposed in order to characterize normal and infarcted myocardium by analysing the two cross-sectional views (parasternal short axis mid-left ventricular view at papillary muscle level and apical four chamber view) of echocardiograms. The Stationary Wavelet Transform (SWT) method is used to extract Relative Wavelet Energy and Entropy (RWE and RWEnt) features from the two cross-sectional views of echocardiography images separately. The proposed method is able to identify MI with an accuracy of 96.80%, sensitivity of 93.70% and specificity of 100% using 16 features extracted from only two frames making this a more reliable classification. An integrated index called Myocardial Infarction Index (MII) is proposed to discriminate infarcted and normal myocardium using features extracted from apical cross-sectional views of echocardiograms. The cross-sectional view of normal and MI echocardiography images is represented as textons using Maximum Responses (MR8) filter banks from which various features are extracted. Then, combinations of highly ranked features are used in the formulation and development of an integrated MII. This calculated novel MII is used to accurately and quickly detect infarcted myocardium by using one numerical value. This technique is able to characterize MI with an average accuracy of 94.37%, sensitivity of 91.25% and specificity of 97.50% with 8 apical four chambers view features extracted from only single frame per patient. Finally, Curvelet Transform and Local Configuration Pattern (LCP) are employed in this thesis for an automated detection of extent of MI (normal, moderately infarcted and severely infarcted myocardium) using 2D echocardiograms. The methodology extracts the LCP features from curvelet transform coefficients of echocardiograms. The developed algorithm showed an accuracy of 98.99%, sensitivity of 98.48% and specificity of 100% for SVM classifier using only six features. Furthermore, an integrated index called Myocardial Infarction Risk Index (MIRI) is developed to detect the normal, moderately and severely infarcted myocardium using a single number. Thus the proposed systems in this thesis may aid the clinicians in faster identification of MI and extent of infarcted myocardium using texture analysis of 2D echocardiograms. Proposed systems may also aid in identifying the person at risk of developing heart failure based on the extent of infarcted myocardium.