Cross-Modality LGE-CMR Segmentation Using Image-to-Image Translation Based Data Augmentation

• Published in: IEEE/ACM transactions on computational biology and bioinformatics Vol. 20; no. 4; pp. 2367 - 2375
• Main Authors: Wang, Wei, Yu, Xinhua, Fang, Bo, Zhao, Yue, Chen, Yongyong, Wei, Wei, Chen, Junxin
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.07.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Bi-ventricular segmentation; Biomedical imaging; cross-modal MRI; Data augmentation; Gadolinium; Image coding; Image processing; Image reconstruction; Image segmentation; image translation; Magnetic resonance; Magnetic resonance imaging; Myocardial infarction; Myocardium; Networks; Training; unsupervised; Ventricle
• ISSN: 1545-5963; 1557-9964; 1557-9964
• DOI: 10.1109/TCBB.2022.3140306

Accurate segmentation of ventricle and myocardium from the late gadolinium enhancement (LGE) cardiac magnetic resonance (CMR) is an important tool for myocardial infarction (MI) analysis. However, the complex enhancement pattern of LGE-CMR and the lack of labeled samples make its automatic segmentation difficult to be implemented. In this paper, we propose an unsupervised LGE-CMR segmentation algorithm by using multiple style transfer networks for data augmentation. It adopts two different style transfer networks to perform style transfer of the easily available annotated balanced-Steady State Free Precession (bSSFP)-CMR images. Then, multiple sets of synthetic LGE-CMR images are generated by the style transfer networks and used as the training data for the improved U-Net. The entire implementation of the algorithm does not require the labeled LGE-CMR. Validation experiments demonstrate the effectiveness and advantages of the proposed algorithm.