PET-MRI Image Fusion Using Adaptive Filter Based on Spectral and Spatial Discrepancy

Recently, medical imaging equipment has undergone major developments. They play an important role in healthcare industry since they provide visual interpretation of human organs. Magnetic resonance imaging (MRI) and positron emission tomography (PET) are two well-known technologies which capture the structural and functional characteristics of the body organs, respectively. Fusing such functional and structural information can help physicians to better understand the normal and abnormal behaviors of tissues and organs. The contribution of this paper is twofold. First, an adaptive filter-based image fusion method is proposed to integrate information of MRI and PET images. Second, the notion of spatial discrepancy is added to the conventional spectral discrepancy and both discrepancy criteria are used for optimizing the filter parameters. The proposed image fusion technique is tested on seven publicly available brain image datasets including two cases of Alzheimer's disease, normal coronal, normal sagittal, normal axial, grade II astrocytoma, and grade IV astrocytoma provided by Harvard University. The proposed method is compared with other biomedical image fusion approaches in both quantitative and visual manners. The results show that the proposed image fusion technique provides better performance in improving spectral and structural characteristics of the original images.