Image Synthesis in Multi-Contrast MRI With Conditional Generative Adversarial Networks

By Salman Ul Hassan Dar, Mahmut Yurt, Levent Karacan, Aykut Erdem, Erkut Erdem, Tolga Cukur. IEEE Transactions on Medical Imaging, 2018

581 citations recorded by the source corpus.

Acquiring images of the same anatomy with multiple different contrasts increases the diversity of diagnostic information available in an MR exam. Yet, the scan time limitations may prohibit the acquisition of certain contrasts, and some contrasts may be corrupted by noise and artifacts. In such cases, the ability to synthesize unacquired or corrupted contrasts can improve diagnostic utility. For multi-contrast synthesis, the current methods learn a nonlinear intensity transformation between the source and target images, either via nonlinear regression or deterministic neural networks. These methods can, in turn, suffer from the loss of structural details in synthesized images. Here, in this paper, we propose a new approach for multi-contrast MRI synthesis based on conditional generative adversarial networks. The proposed approach preserves intermediate-to-high frequency details via an adversarial loss, and it offers enhanced synthesis performance via pixel-wise and perceptual losses for registered multi-contrast images and a cycle-consistency loss for unregistered images. Information from neighboring cross-sections are utilized to further improve synthesis quality. Demonstrations on T1- and T2- weighted images from healthy subjects and patients clearly indicate the superior performance of the proposed approach compared to the previous state-of-the-art methods. Our synthesis approach can help improve the quality and versatility of the multi-contrast MRI exams without the need for prolonged or repeated examinations.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/757993ad8c3574857a01906d2b46a5909882a4fa

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Sergey I. Nikolenko
2. Ling Shao
3. Jianbing Shen
4. Huan Fu
5. Tao Zhou
6. Wenguan Wang
7. Deng-Ping Fan
8. Geng Chen

Citations in the corpus

• Generative Adversarial Network in Medical Imaging: A Review (2018, methodology): “’s work is similar to [30], where transformations were across T1, T2 and T2-Flair MR images [158].”
• ResViT: Residual Vision Transformers for Multimodal Medical Image Synthesis (2021, methodology): “Trained with image-average loss terms, CNNs have difficulty in coping with atypical anatomy that substantially varies across subjects [24], [43].”
• Computer‐aided diagnosis of prostate cancer using a deep convolutional neural network from multiparametric MRI (2018, methodology): “Generative adversarial networks could be another choice to synthesize mp-MR images to increase the number of dataset samples.(25) In conclusion, we have proposed a patch-based DCNN model to classify prostate PCa and NC from mpMR images.”
• A Review of Deep-Learning-Based Medical Image Segmentation Methods (2021, methodology): “[64] proposed a new method of multicontrast MRI synthesis based on conditional generative adversarial networks.”
• Hi-Net: Hybrid-Fusion Network for Multi-Modal MR Image Synthesis (2020, methodology): “In addition, GAN-based methods have also achieved promising results in synthesizing various types of medical images, such as CT images [15], retinal images [34], [35], MR images [19], ultrasound images [36], [37] and so on.”
• Generative artificial intelligence: a systematic review and applications (2024, unknown): “Furthermore, Dar et al. [28] used Conditional GANs to solve the problem of Image translation in MRI.”
• Medical Image Synthesis for Data Augmentation and Anonymization using Generative Adversarial Networks (2018, methodology): “GANs have been used to generate a motion model from a single preoperative MRI [10], upsample a lowresolution fundus image [13], create a synthetic head CT from a brain MRI [16], and synthesizing T2-weight MRI from T1-weighted ones (and vice-versa) [5].”
• MedGAN: Medical Image Translation using GANs (2018, methodology): “Other medical translation tasks have also been recently explored such as CT to PET (Ben-Cohen et al., 2018) and 2T MR to 1T MR translation (Yang et al., 2018b; Dar et al., 2018).”
• Synthetic Data for Deep Learning (2019, background): “H&E staining, including transformation from unstained to stained lung histology images [43] and stain style transfer [538], multi-contract MRI (from contrast to contrast) [138], 3D cross-modality MRI [677], different styles of prostate histopathology [488], different datasets of chest X-rays [98], a”