Research On Some Problems Of Multimodal Medical Image Fusion

Image fusion technology can effectively combine the salient information of multiple images and express it through one image to break through the limitations of insufficient information transmission by a single imaging mechanism.In the field of medical imaging,there are many different medical imaging devices such as computer tomography(CT)and magnetic resonance imaging(MRI)based on different imaging principles.The fused multimodal images can more comprehensively describe the health status of the patient's organs and provide a wealth of auxiliary information for doctors to develop more suitable treatment plans.The fusion quality of the multimodal medical image mainly depends on the two modules of image decomposition and image fusion.Among them,the separation effect of the base information and texture details of the source image directly affects the result of the fusion module.Specifically,the more thoroughly the base information and texture details are separated,the more beneficial it is to improve the quality of the fused image.Therefore,in the traditional image fusion framework,the multimodal medical image fusion methods based on multi-scale decomposition have achieved relatively excellent fusion effects,but there are still some problems to be solved,such as: the inconsistency of the decomposition base of the multi-scale decomposition method;the type of multi-scale features of the image is single;the image decomposition method lacks adaptability,etc.This paper aims to further improve the quality of the fused image,and study the multimodal medical image fusion algorithms from three aspects.The specific research contents are as follows:1.MRI and CT image fusion based on synchronized-anisotropic diffusion modelIn view of the inconsistency of the source image decomposition base in the image decomposition method,a fusion method for MRI and CT images based on synchronized-anisotropic diffusion equation(S-ADE)is proposed.First,the modified S-ADE model is used to decompose the MRI and CT images to obtain the base layer and texture layer of the corresponding source image.Then,the "maximum absolute value" rule is used to fuse the information of the base layers.On texture layers,the fusion decision map is calculated by new sum of modified anisotropic Laplacian(NSMAL)algorithm,which is designed using the common decomposition coefficients obtained by S-ADE.A consistency check algorithm is designed for the obtained texture layer decision map to mitigate the staircase effects in the fused image.Then,the fused base layer and the fused texture layer are combined by linear addition to obtain the fused image.Finally,the fused MRI-CT image is obtained after image correction.Experimental results show that the S-ADE method can retain a large amount of useful information of the source images while guaranteeing image quality and visual effects.Based on the subjective and objective evaluation results,the S-ADE algorithm performs better than other general fusion algorithms.2.MRI and CT image fusion based on hybrid feature decompositionIn order to break through the limitations of the decomposition algorithms based on the single-category features of the image,a decomposition method that effectively utilizes the hybrid features of the image is proposed.This method combines the advantages of feature representation methods based on spatial domain and transform domain.It can effectively separate the base information and texture details of the image,which is conducive to the better effect of the fusion rules,and improves the execution efficiency of the transform domain-based algorithm.First,the source anatomical image is decomposed into a series of high frequencies and a low frequency via nonsubsampled shearlet transform(NSST)method.Second,the low frequency is further decomposed using the designed optimization model based on structural similarity and structure tensor to get an energy texture layer and a base layer.Then,the modified choosing maximum(MCM)algorithm is designed to fuse base layers.The sum of modified Laplacian(SML)is used to fuse high frequencies and energy texture layers.The fused low frequency can be obtained by adding fused energy texture layer and base layer.Finally,the fused image is obtained through the inverse NSST.The superiority of this method is verified through experiments on 50 pairs of MRI/CT images and others.Compared with 12 general medical image fusion methods,experiments show that the proposed hybrid decomposition model has a better ability to extract texture information than traditional methods.3.Multimodal medical image fusion using adaptive co-occurrence filterbased decomposition optimization modelIn order to solve the problem that filter-based image fusion methods cannot adaptively process medical images and require multiple iterations,the concept of the skewness of pixel intensity(SPI)and an adaptive co-occurrence filter(ACOF)-based image decomposition optimization model are proposed.First,the pixel intensity skew value of the source image is calculated,based on which the scale parameters of the cooccurrence filter are calculated,and an adaptive co-occurrence filter is obtained.The filter can be executed according to the content of the image,rather than relying on a fixed filter scale.The source image is filtered by ACOF to obtain the initial base layer of the image.Then,an optimization model is constructed based on the characteristics of the pixel intensity distribution of the base layer to replace the widely used iterative filter framework to obtain the final base layer,ensuring the complete separation of the base information and texture details of the image.Finally,the fused image is generated according to the designed fusion rules.Experimental results show that the ACOF method has outstanding performance in six objective indicators and subjective evaluation,and it has higher computational efficiency.The main innovations are summarized as follows: the problem of different source image decomposition bases is solved,a consistent image decomposition-fusion framework is constructed,and the phenomenon of residual texture in the fused image is solved through the consistency check algorithm and the image correction method;a three-scale image decomposition model based on hybrid features of image is proposed by combining the advantages of spatial domain and transform domain-based methods to improve the efficiency of algorithm while ensuring the quality of the fused image;the proposed adaptive co-occurrence filter decomposition optimization algorithm solves the problem of iterative execution of the filters and makes the image decomposition algorithm adaptive.