基于深度学习的医学图像合成：方法、应用与未来方向
DOI:
https://doi.org/10.70693/cjmsr.v1i3.1764Keywords:
深度学习, 医学图像合成, 生成对抗网络, 扩散模型, 跨模态合成Abstract
随着医学影像技术的快速发展，深度学习在医学图像合成领域展现出巨大潜力，为临床诊断、治疗规划和医学研究提供了新机遇。本文系统综述了基于深度学习的医学图像合成方法，重点探讨了生成对抗网络（GANs）在跨模态合成和域适应中的应用，扩散模型（DM）的高保真度生成能力及其在多模态合成中的局限性（如合成样本多样性不足），以及Transformer模型和混合模型与领域适应技术的优势。在应用场景方面，本文分析了跨模态图像合成（例如MRI到CT、MRI到PET转换）、模态内图像增强与标准化、图像重建与去噪、纵向图像生成与疾病进展建模、特定病变或组织合成等实践案例，同时指出现有方法在泛化性、计算效率和伦理合规性方面的局限。未来研究应聚焦于提升模型的可解释性与可靠性、多模态融合能力及临床应用可行性。
References
[1] Ozbey M, Dalmaz O, Dar SUH, et al. Unsupervised Medical Image Translation With Adversarial Diffusion Models[J]. IEEE Trans Med Imaging,2023,42(12):3524-3539.
[2] Emami H, Dong M, Nejad-Davarani SP, et al. Generating synthetic CTs from magnetic resonance images using generative adversarial networks[J]. Med Phys. 2018;45: 3627-3636.
[3] Ozonoff A, Schaenman J, Jayavelu ND, et al. Phenotypes of disease severity in a cohort of hospitalized COVID-19 patients: Results from the IMPACC study[J]. EBioMedicine. 2022;83:104208.
[4] Guogang Cao, Shunkun Liu, Hongdong Mao, et al. Improved CyeleGAN for MR to CT synthesis[C/OL]//2021 6th International Conference on Intelligent Informatics and Biomedical Sciences (ICIIBMS)[J]. IEEE: 2021;205-208.
[5] Lin W, Lin W, Chen G, et al. Alzheimer’s Disease Neuroimaging Initiative. Bidirectional Mapping of Brain MRI and PET With 3D Reversible GAN for the Diagnosis of Alzheimer's Disease[J]. Front Neurosci. 2021;15:646013.
[6] Vasylechko S, Afacan O, Kurugol S. Self Supervised Denoising Diffusion Probabilistic Models for Abdominal DW-MRI[J]. Comput Diffus MRI. 2023;14328:80-91.
[7] Xie T, Cao C, Cui ZX, et al. Synthesizing PET images from high-field and ultra-high-field MR images using joint diffusion attention model[J]. Med Phys. 2024;51(8):5250-5269.
[8] Dalmaz O, Yurt M, Cukur T. ResViT: Residual Vision Transformers for Multimodal Medical Image Synthesis[J]. IEEE Trans Med Imaging. 2022;41(10):2598-2614.
[9] Li Y, Zhou T, He K, et al. Multi-Scale Transformer Network With Edge-Aware Pre-Training for Cross-Modality MR Image Synthesis[J]. IEEE Trans Med Imaging. 2023;42(11):3395-3407.
[10] Wei W, Poirion E, Bodini B, et al. Predicting PET-derived demyelination from multimodal MRI using sketcher-refiner adversarial training for multiple sclerosis[J]. Med Image Anal. 2019;58:101546.
[11] Zhang L, Xiao Z, Zhou C, et al. Spatial adaptive and transformer fusion network (STFNet) for low-count PET blind denoising with MRI[J]. Med Phys. 2022;49(1):343-356.
[12] Liu X, Xing F, Prince JL, et al. Dual-cycle constrained bijective VAE-GAN for TAGGED-TO-CINE magnetic resonance image synthesis [J]. Proc IEEE Int Symp Biomed Imaging. 2021; 10:1109
[13] Yan S, Wang C, Chen W, et al. Swin transformer-based GAN for multi-modal medical image translation[J]. Front Oncol. 2022; 12:942511.
[14] Zhou T, Fu H, Chen G, et al. Hi-Net: Hybrid-Fusion Network for Multi-Modal MR Image Synthesis[J]. IEEE Trans Med Imaging. 2020 ;39(9):2772-2781.
[15] Li W, Kazemifar S, Bai T, et al. Synthesizing CT images from MR images with deep learning: model generalization for different datasets through transfer learning[J]. Biomed Phys Eng Express. 2021 ;24;7(2).
[16] Klages P, Benslimane I, Riyahi S, et al . Patch-based generative adversarial neural network models for head and neck MR-only planning[J]. Med Phys. 2020 ;47(2):626-642.
[17] Hsu SH, Han Z, Leeman JE, et al. Synthetic CT generation for MRI-guided adaptive radiotherapy in prostate cancer[J]. Front Oncol. 2022;23;12:969463.
[18] Shin, H.C., Ihsani,et al. Generative adversarial networks with bidirectional encoder representations from transformers for mri to pet synthesis.2020;2008:04393.
[19] Lin H, Figini M, D'Arco F, et al. Low-field magnetic resonance image enhancement via stochastic image quality transfer[J]. Med Image Anal. 2023 ;87:102807.
[20] Liu J, Pasumarthi S, Duffy B, et al. One Model to Synthesize Them All: Multi-Contrast Multi-Scale Transformer for Missing Data Imputation[J]. IEEE Trans Med Imaging. 2023;42(9):2577-2591.
[21] Schwab J, Antholzer S, Haltmeier M. Big in Japan: Regularizing Networks for Solving Inverse Problems[J]. J Math Imaging Vis. 2020;62(3):445-455.
[22] Häggström I, Schmidtlein CR, Campanella G, et al. DeepPET: A deep encoder-decoder network for directly solving the PET image reconstruction inverse problem[J]. Med Image Anal. 2019 ;54:253-262.
[23] Wang X, Yang M, Tosun S, et al. CalDiff: calibrating uncertainty and accessing reliability of diffusion models for trustworthy lesion segmentation[J]. IEEE J Biomed Health Inform. 2025;23;PP.
[24] Eshraghi Boroojeni P, Chen Y, Commean PK, et al. Deep-learning synthesized pseudo-CT for MR high-resolution pediatric cranial bone imaging (MR-HiPCB)[J]. Magn Reson Med. 2022;88(5):2285-2297.
[25] Qi M, Li Y, Wu A, et al. Multi-sequence MR image-based synthetic CT generation using a generative adversarial network for head and neck MRI-only radiotherapy[J]. Med Phys. 2020;47(4):1880-1894.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 杨莉 (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
