基于多模态图像融合的腰部脊柱MRI图像分割研究OA

Objective To propose a multimodal magnetic resonance imaging(MRI)image fusion and segmentation method,and combine different sequence images under MRI to improve the segmentation accuracy of organs in the spinal region.Methods A dual-branch fusion segmentation network was designed for T1 and T2 modal images under MRI.Firstly,the features in a single modality were extracted by using a hybrid encoder.Then,the cross-modal features were extracted through the Transformer structure.Finally,in the upsampling stage,the multi-modal information was fused based on the spatial attention mechanism to complete the segmentation.Furthermore,a deep supervision mechanism was adopted to improve the efficiency of network training.Experiments were conducted on the lumbar spine MRI dataset SPIDER.The method proposed in this paper was compared with the traditional single-modal segmentation method and other fusion segmentation methods to verify the effectiveness of feature fusion of the method proposed in this paper.Results The Dice coefficients of the proposed method in this paper for segmentation on vertebrae,intervertebral discs and spinal canals were 93.20%,86.90%and 94.80%respectively,which were superior to the traditional single-modal image segmentation methods.Moreover,the proposed method in this paper was more robust in the case of modal absence.Conclusion The multimodal MRI image fusion and segmentation method proposed in this paper can effectively combine the respective imaging advantages of the T1 and T2 modalities,improve the segmentation accuracy of organs in the spinal region,assist doctors in diagnosis in clinical applications,and improve the efficiency of diagnosis and treatment.