Journal of Cardiovascular Disease Research
MACHINE LEARNING FRAMEWORK FOR MEDICAL IMAGE FUSION APPROACH
N. Karthika , Manupati Shruthi , Maripelli Rishitha , Mogili Poojitha , Kuthadi Keerthi
JCDR. 2023: 638-648
Abstract
Fusion process gives highly informative image as it combines the information from two or more images into a single image. It has been utilizing widely in medical research field for computer aided brain surgery, Alzheimer’s treatment, tumour detection and other clinical diagnosis. Effective fusion algorithms are required to obtain accuracy of successful diagnosis of diseases. Magnetic resonance (MR) and computed tomography (CT) images are most widely utilized images for analysing the human body. The main objective of any fusion approach is to transfer maximum information from the source images to the fused image with minimum information loss. It must minimize the artifacts in the fused image. In this work, a novel medical image fusion algorithm is proposed. Nonlinear anisotropic filtering (NLAF) in principal component analysis (PCA) domain, which preserve the texture information of fused images most effectively. NLAF is utilized to decompose the source images into approximation and detail layers. Final detail and approximation layers are computed with the support of PCA. Finally, fused image is generated from the linear combination of final detail and approximation layers. Qualitative and quantitative performance of the proposed algorithm is assessed with the help of image quality metrics like peak signal-to-noise ratio (PSNR), correlation coefficient (CC), entropy (E), root mean square error (RMSE) and structural similarity (SSIM) index. Extensive simulation results of the proposed hybrid algorithm are compared with the traditional and recent image fusion algorithms. Performance evaluation discloses that the proposed fusion approach outperforms the existing fusion method
» PDF