Automatic cardiac MRI localization method

Cardiovascular diseases remain the biggest cause of death worldwide. Early detection is the most important step to solve this problem. Cardiac Magnetic Resonance Imaging (CMR) has shown its capability of imaging the heart and evaluation of cardiac function without exposure to ionizing radiation. In this paper, we aim to perform automatic heart localization in MRI short axis view using a simple and effective way. The proposed algorithm can be described through the following two steps: (a) preprocessing step to denoise the images using Gaussian filter, and (b) localization step based on shape recognition technique. The algorithm, implemented using MATLAB, was developed and tested using two datasets. First one consists of 33 subjects normal and abnormal for a total of 7980 2D images and the second one consists of 50 2D images related to normal subject. The results here are evaluated through the classification into three types: totally localized case (complete heart falls inside the selected ROI window), partially localized case (small part, less than 10%, of LV or RV falls outside the selected ROI window), and not localized case (more than 10% of the hearts falls outside the selected ROI window). It gives result with 91.4% accuracy for totally localized case, 8.1% for partially localized case, and 0.5% for not localized case.