Feature Level Medical Image Fusion with Deep Learning

In recent years, medical image fusion has emerged as a comprehensive analytical method that frequently employs various modification techniques. The objective of multi-modal medical image fusion is to enhance image quality while preserving specific characteristics by combining images from different imag- ing modalities. This field encompasses cutting-edge domains such as image processing, computer vision, pattern recognition, ma- chine learning, and artificial intelligence. Furthermore, medical image fusion has found extensive clinical applications, enabling healthcare professionals to gain a deeper understanding of lesions by integrating medical images from multiple modalities. In this study, we propose the utilization of a Grey Wolf algorithm, which comprises various learning models, to minimize squared error and optimize the process through weighted entropy during image fusion. The successful results are attributed to a hybridization of the Discrete Wavelet Transform (DWT) and the Discrete Curvelet Transform (DCT). By employing Principal Component Analysis, Grey Wolf Optimization, and a Recurrent Neural Network, we have developed a novel fusion model for unsupervised categorization known as GWO-RNN. This model is designed to work with multi-modality images that incorporate structural and sparse constraints. The flexibility and precision of image fusion are further enhanced through correlation and covariance analysis using eigenvalues and eigenvectors. Matlab simulations of the proposed workbench are conducted, comparing it against state-of-the-art methods to assess performance metrics, including peak signal-to-noise ratio, mean square error, correlation, and accuracy. The proposed workbench not only meets but surpasses all performance benchmarks while also reducing computational overhead.