Comprehensive performance analysis of different medical image fusion techniques for accurate healthcare diagnosis applications

The advancement of medical imaging has led to the acquisition of image data from multiple modalities, necessitating the development of robust algorithms for accurate and reliable fusion of such diverse image sets. Medical image fusion plays a crucial role in enhancing the clinical applicability of medical images by combining information from different modalities into a single fused image that provides comprehensive and instructive insights. In recent years, significant efforts have been devoted to expanding the repertoire of image fusion algorithms, particularly in the absence of standardized benchmarks and comprehensive code libraries that can support state-of-the-art techniques. This research presents a comprehensive performance analysis of different medical image fusion techniques applied to a wide range of medical images. To facilitate this analysis, we have curated a Medical Image Fusion Benchmark (MIFB) that incorporates a diverse set of evaluation metrics (EM) for quantitative assessment. Additionally, we have developed a Fusion Algorithms Code Library (FACL) that provides a comprehensive repository of fusion algorithms for easy access and comparative analysis. Through rigorous experiments conducted within this benchmarking framework, we aim to identify the most effective algorithms for achieving powerful image fusion, considering both quantitative and qualitative outcomes. Moreover, we offer insightful observations on the current state and future prospects of the field. To validate the effectiveness of our approach, we compare our results with previous related work in the field. Our comparative analysis reveals that Principal Component Analysis (PCA) demonstrates superior performance in fusing MRI and CT medical images, exhibiting a restoration quality with average PSNR values of 19.047 dB and RMSE of 6.1 x 10(-4). On the other hand, Convolutional Sparse Representation (Conv_SR) outperforms other techniques in fusing MRI and PET medical images, achieving average PSNR values of 19.073 dB and RMSE of 5.51 x 10(-4), indicating good restoration quality. Lastly, the Non-Subsampled Shearlet Transform_Parameter-Adaptive Pulse-Coupled Neural Network (NSST_PA-PCCN) attains impressive performance in fusing MRI and SPECT medical images, with average PSNR values of 19.101 dB and RMSE of 4.90382 x 10(-4), suggesting high-quality restoration. In conclusion, this research contributes to the comprehensive analysis and evaluation of various medical image fusion techniques for accurate healthcare diagnosis applications. By establishing the Medical Image Fusion Benchmark and Fusion Algorithms Code Library, we offer a valuable resource for researchers and practitioners in the field. Our findings highlight the superior performance of specific techniques, such as PCA, Conv_SR, and NSST_PA-PCCN, for fusing different medical image modalities and achieving excellent restoration quality. The insights gained from this study can guide the selection and implementation of appropriate image fusion algorithms in medical imaging applications, ultimately contributing to improved healthcare diagnostics.