摘要
Medical image fusion improves clinical interpretation and analysis by combining the complementary information of multimodal images into one that leads to more accurate diagnosis and treatment planning. This article presents a novel feature-level medical image fusion (FMIF) method using a structural gradient-based decomposition, which provides uncorrelated structural and textural components. A feature codebook obtained from multiple low-scale features followed by clustering and choose-max with consistency verification rule is applied to fuse structural components. An optimized pulse-coupled neural network is utilized to fuse texture component using a differential evolution algorithm, which helps to improve the model efficiency and retain the natural response of pixel activity. The fusion performance of the proposed method is explored on a large data set of neurological images. The experimental result demonstrates that the proposed method provides better fusion results and outperforms the state-of-the-art fusion approaches with enhanced visual quality and computational parameters.
学科与关键词
- Editorial: Multimodal brain image fusion: Methods, evaluations, and applicationsLiu, Yu*; Ma, Jiayi*; Zhang, Qiang*; Wei, Wei*; Chen, Xun*; Liu, Zheng*Frontiers in Neuroscience, 2023, 16: 1128938.
- Multimodal medical image fusion review: Theoretical background and recent advancesHermessi, Haithem*; Mourali, Olfa; Zagrouba, EzzeddineSIGNAL PROCESSING, 2021, 183: 108036.被引 151
- Multimodal Medical Image Fusion using Rolling Guidance Filter with CNN and Nuclear Norm MinimizationLiu, Shuaiqi; Yin, Lu; Miao, Siyu; Ma, Jian; Cong, Shuai*; Hu, ShaohaiCURRENT MEDICAL IMAGING, 2020, 16(10): 1243-1258.被引 16
