To address the issues of information integrity loss and insufficient cross-modal feature extraction in medical image fusion methods, a novel medical image fusion approach based on an improved U-Net and cross-modal self-distillation is proposed. This method enhances the encoding part of U-Net by designing a dual-branch encoder that combines the strengths of CNN and Transformer, enabling more effective capture and retention of local and global features in medical images, thus solving the problem of information integrity loss. Cross-modal self-distillation technology is employed to facilitate information exchange between the CNN branches and the Transformer branches of two medical images, enhancing the interaction between different modal features and maximizing the acquisition of cross-modal features. In the decoding stage, an Attention Gate mechanism is proposed to replace the skip connections in U-Net, ensuring that the network can effectively focus on key features and further enhance the integrity of the fused images. Experimental results show that, compared to other methods, the fusion images obtained by this method not only preserve more complete texture details and edge information but also effectively solve the problem of insufficient cross-modal feature extraction.
VESHKIF G, OUZIRN, VOROBYOVS A, et al.Coupled feature learning for multimodal medical image fusion[J/OL]. arXiv: 2102.08641, 2021vl.
[2]
PALKARB, MISHRAD. Fusion of multi-modal l-umbar spine images using Kekre’s hybrid wavelet t-ransform[J]. IET Image Processing, 2019, 13(12): 2271-2280.
[3]
YANGY. Multimodal medical image fusion through a new DWT based technique[C]//2010 4th Internat-ional Conference on Bioinformatics and Biomedical Engineering, 2010: 1-4.
[4]
CUNHA A LDA, ZHOUJ, DOM N. The nonsu-bsampled contourlet transform: theory, design, and a-pplications[J]. IEEE Transactions on Image Processi-ng, 2006, 15(10): 3089-3101.
[5]
LIUY, LIUS, WANGZ. A general framework for image fusion based on multi-scale transform and sparse representation[J]. Information Fusion, 2015, 24: 147-164.
[6]
ZHANGY, YANGM, LIN, et al. Analysis-synth-esis dictionary pair learning and patch saliency me-asure for image fusion[J]. Signal Processing, 2020, 167: 107327.
[7]
LIH, WUX J. Multi-focus image fusion using d-ictionary learning and low-rank representation[C]//International Conference on Image & Graphics.Cham: Springer, 2017: 675-686.
[8]
LIUY, CHENX, CHENGJ, et al. A medical image fusion method based on convolutional neural networks[C]//2017 20th International Conference on I-nformation Fusion, 2017: 1-7.
[9]
JIANL, YANGX, LIUZ, et al. SEDRFuse: a symmetric encoder-decoder with residual block network for infrared and visible image fusion[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 70: 1-15.
[10]
ZHAOZ, BAIH, ZHANGJ, et al. CDDFuse: Co-rrelation-driven dual-branch feature decomposition for multi-modality image fusion[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 5906-5916.
[11]
DINHL, SOHL-DICKSTEINJ, BENGIOS. Den-sity estimation using Real NVP[J/OL]. [2024-02-18].
[12]
XUEZ, GAOZ, RENS, et al. The modality fo-cusing hypothesis: towards understanding crossmodal knowledge distillation[J/OL]. [2024-02-18].
[13]
ZHOUS, LIUW, HUC, et al. UniDistill: a un-iversal cross-modality knowledge distillation framework for 3D object detection in bird's-eye view[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 5116-5125.
TIANJialin, XUXing, SHENFumin, et al. Cross-modal self-distillation for zero-shot sketch-based i-mage retrieval[J]. Journal of Software, 2022, 33(9): 3152-3164. (in Chinese)
[16]
YUW, ZHOUP, YANS, et al. Inceptionnext: when inception meets ConvNeXT[J/OL]. [2024-02-18].
[17]
ZHUL, WANGX, KEZ, et al. Biformer: vision transformer with bi-level routing attention[C]//Proc-eedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 10323-10333.
XIENing, CHENLiang, PEIZiqing, et al. Weldingspot positioning method for body in white based o-n improved U-Net[J]. Journal of Test and Measure-ment Technology, 2024, 38(1): 12-18. (in Chinese)
[20]
WOO S, PARKJ, LEEJ Y, et al. CBAM: convo-lutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 3-19.
[21]
ZHANGY, LIUY, SUNP, et al. IFCNN: a general image fusion framework based on convolutional neural network[J]. Information Fusion, 2020, 54: 99-118.
[22]
XUH, MAJ, JIANGJ, et al. U2Fusion: a unified unsupervised image fusion network[J]. IEEE Tra-nsactions on Pattern Analysis and Machine Intellige-nce, 2022, 44(1): 502-518.
[23]
TANGW, HEF, LIUY, et al. MATR: multimod-al medical image fusion via multiscale adaptive tra-nsformer[J]. IEEE Transactions on Image Processing, 2022, 31: 5134-5149.
[24]
FUJ, LIW, DUJ, et al. A multiscale residual p-yramid attention network for medical image fusion[J]. Biomedical Signal Processing and Control, 2021, 66: 102488.
[25]
WENJ, QINF, DUJ, et al. Msgfusion: medical semantic guided two-branch network for multimodal brain image fusion[J]. IEEE Transactions on Multimedia, 2023, 26: 944-957.