A Systematic Literature Review on Multimodal Medical Image Fusion

Medical image fusion is a relevant area with widespread application in disease diagnosis and prediction with easily available image scans of Computed Tomography, Positron Emission Tomography, Magnetic Resonance Imaging, and Single Photon Emission Computed Tomography. Each diagnostic image modality has its advantages and limitations. Multimodal Medical Image Fusion aims to combine more than one image of the same or different modality to enhance the image content and provide more information about diseases. We performed a Systematic Literature Review according to the methodology outlined in Kitchenham Charter and based on our search string, we extracted 844 studies from four electronic databases published between 2017 and 2021. Around 175 studies were selected for further in-depth analysis using inclusion and exclusion criteria. We further divide this review article into five sections that (a) Identify the most frequently used input image decomposition methods (b) Describes the most common fusion rules applied on decomposed sub-bands (c) Discusses the optimization algorithms used to improve the efficiency of the fusion scheme (d) Examines the modalities which are subjected to image fusion techniques in the medical domain (e) Identifies the evaluation metrics used to judge the effectiveness of image fusion technique. The result of the comparative study of five sections highlights that the majority of studies use multiscale decomposition methods, and hybrid and neural network-based fusion rules while the CT-MRI combination was mostly used as an input dataset. The review also indicated the prevalent use of particle swarm optimization and non-reference metrics in the majority of studies. Our results suggest that medical image fusion can improve the quality and accuracy of medical images for diagnosis and treatment planning. Further research can be conducted to handle potential research gaps outlined in this review and optimize medical image fusion for clinical applications.