Ventricular analysis of cardiac magnetic resonance images using automated hybrid segmentation methods

Abstract

Ventricular hypertrophy is one among the prevalent cardiovascular diseases for which Cardiac Magnetic Resonance (CMR) images are referred for examining the cardiac morphology and its function. Manual delineation of the ventricles is time consuming and there is a probability of having an intra-observer and inter-observer variability. Hence there is a need to automate the diagnostic process to improve the sensitivity and accuracy of the test. The objective of this work is to develop hybrid segmentation methods for automatic segmentation of both left and right ventricles from intensity-inhomogeneity CMR images. Analysis is carried out on 35 data sets consisting of both normal and abnormal short-axis cine CMR images recorded under a steady state free precision protocol obtained from 3T Magnetic Resonance Imaging (MRI) unit (Magnetom Symphony, Siemens Medical Solutions, Erlangen, Germany) of Rajiv Gandhi Government General Hospital, Park Town, Chennai. Image segmentation is an important step in medical image analysis and segmentation of ventricles in CMR images is challenging due to an inbuilt artifact called intensity-inhomogeneity. The short axis cine CMR images were corrected for intensity-inhomogeneity using Bias Corrected Fuzzy Cmeans Method (BCFCM), Level Set (LS) and Multiplicative Intrinsic Component Optimization (MICO) methods. The average value of skewness for BCFCM is o.5 whereas it is 0.8 for original and MICO methods, LS method it is 0.9. The value of skewness near to zero indicates more symmetric data distribution. The statistical measures and multifractal analysis show that bias correction by BCFCM has better performance than MICO and LS. newline newline newline