Left atrium MRI image segmentation using efficient Xception stochastic depth based generative adversarial network

Abstract

ABSTRACT Globally, cardiovascular diseases are an important cause of death. The general underlying mechanism of the key cardiovascular diseases is thrombus formation inside the blood vessels. For these thrombi, the Left Atrial Appendage (LAA) is the significant repository. It is a residual appendix as of the Left Atrium (LA) embryonic development; in addition, thrombus formation in healthy patients is prevented by its high contractility. Hence, LA segmentation is required. Proposing an effectual technique to attain automatic segmentation of the LA of a Magnetic Resonance Imaging (MRI) input image is the goal. By employing a Bitwise Left and Right Shift-centric White Shark Optimizer (BLRSWSO), optimized boundary detection is conducted. To segment LA, the Xception Stochastic Depth-centric Generative Adversarial Network (XSDGAN) is developed. In the end, by deploying the Union Histogram Intersection Box filter (UHIBF), contrast enhancement is carried out. For proving the efficacy, the proposed methodologies’ performance is weighed against the prevailing techniques. The proposed system segments the LA effectively as per the experimentation.