CMMCSegNet: Cross-Modality Multicascade Indirect LGE Segmentation on Multimodal Cardiac MR

Figure 1

The proposed cross-modality multicascade SegNet flowchart of indirect segmentation for LGE modality. bSSFP images are important in distinguishing the cardiac structure from all enhanced areas. Accordingly, the two imaging modalities are treated differently in the proposed method. Cross-modality translation from LGE to bSSFP is chosen as Stage I of the proposed architecture having a greater impact on the LGE segmentation results. bSSFP is regarded as the assistant modality completing the LGE segmentation.

(a) The proposed CMMCSegNet framework

(b) The Cycle-GAN architecture

(c) The multicascade pix2pix architecture