Svm Based Hypertrophy Detection of Compressed CMR Images

Abstract

Ventricular hypertrophy is diagnosed using Cardiac Magnetic Resonance (CMR) images which give important information on the function of ventricles and their size. Ventricular analysis needs exact delineation of the ventricles and the current manual clinical practice involves intra and inter-observer variability’s. Due to huge amount of data that is to be analyzed, it is a tedious and time consuming process. Image compression is a process of reducing or eliminating redundant or irrelevant data from the image so that only essential information can be stored to reduce the storage size, transmission bandwidth and transmission time. Discrete Wavelet Transform (DWT) based method for compression, while effective, are lossy. The wavelet Transform presents a restricted Multi-resolution portrayal of a sign both in time and recurrence. DWT suffer from some loss of information and outcome of image quality is not good. The Discrete Cosine Transform (DCT) is a strategy for changes a sign or image from spatial domain to frequency component. It is widely used technique in image compression. CMR images are preprocessed for intensity in-homogeneity correction and the corrected images are segmented using proposed automated hybrid Conditional Spatial Fuzzy C-Means (CSFCM) with optimization. Classifiers such as Support Vector Machine (SVM), Self Organized Map (SOM) and Bayesian Classifier (BC) are used for classification. A maximum efficiency was achieved using SVM compared to other classifiers for classifying the abnormality of the ventricles. It can be inferred from the results that ventricular hypertrophy can be detected with higher classification accuracy using selection of appropriate features.

Keywords: Cardiac Magnetic Resonance, Ventricular hypertrophy, Compression, Segmentation, Classifier.