Heuristics based Segmentation of Left Ventricle in Cardiac MR Images

In the development of clinical decision supporting system, the clinical image classification and disease prognosis with human organ plays a main role in health care systems. Especially in cardiac Magnetic Resonance Imaging (MRI) diagnosing abnormal condition will be very difficult. Segmentation of left and right ventricles predicts heart problems and abnormalities and can be helpful to diagnose various heart issues. Cardiovascular disease plays a vital role in humans’ life. Diagnosis and prevention of left ventricular segmentation acts as an important part. Segmenting left ventricular is the complex and significant task to intensity and shape similarity with other organs in our body. In our proposed can be used to prevent cardiac disease effectively by automatically segmenting left ventricle with the help of MRI images, Improvised Convolution network and heuristic algorithm to detect the disease with high accuracy. The network is trained by considering the left ventricle’s comparatively small percentage of pixels in the overall image, . In the post- processing stage, the regions are determined based on their shapes and thresholding on the result image of the fully convolutional network. The performance is measured using the accuracy and loss observed as 96.79 percentage and 0.0029. The input image size was processed to 256 x 256 and the mask fitted the accuracy parameters at this size to the optimized rate of result.