Certain investigations on cardiac masses prediction in echocardiogram images using soft computing techniques

Abstract

Intracardiac masses identification in the images of echocardiogram newlineimages is one of the most essential tasks in making the diagnosis of cardiac newlinedisease. For making the improvement in accuracy over the diagnosis, a new newlinecomplete method of classifying the echocardiogram images automatically newlinewhich is based on double convolutional neural network algorithm. Initially, newlinecropping of images over the specific region is done in order to make the newlinedefinition of the mass area. Later on, as the second step, the processing of newlineglobally unique denoising technique is being implied for the removal of newlinespeckle and in order to make the preservation of anatomical structured newlinecomponent in the image. A robust back propagation neural network (RBPNN) newlinetechnique is used to conquer every single conventional-issue subjected to a newlinetotal of 108 real time dataset of clinical echocardiogram images. It consists newlinefour phases such as noise removal, automatic segmentation, feature newlineextraction, and intracardiac masses classification. Initially, the noise is newlinediminished from the echocardiogram images utilizing the adaptive vector newlinemedian filter (AVMF). Then, linear iterative vessel segmentation (LIVS) is newlineapplied for automatic segmentation of the masses followed by the extraction newlineof 11 best features extracted using the multiscale local binary pattern (MSLBP) newlineapproach. Finally, RBPNN is employed to classify the cardiac masses newlinefrom echocardiogram images. The proposed AVMF-MS-LBP based RBPNN newlineapproach can help the radiologists to diagnose and automatically classify the newlineheart diseases from the echocardiogram images. In newline