Super-resolution is widely used in medical imaging to enhance low-quality data, reducing scan time and improving abnormality detection. Conventional super-resolution approaches typically rely on paired datasets of downsampled and original high resolution images, training models to reconstruct high resolution images from their artificially degraded counterparts. However, in real-world clinical settings, low resolution data often arise from acquisition mechanisms that differ significantly from simple downsampling. As a result, these inputs may lie outside the domain of the training data, leading to poor model generalization due to domain shift. To address this limitation, we propose a distributional deep learning framework that improves model robustness and domain generalization. We develop this approch for enhancing the resolution of 4D Flow MRI (4DF). This is a novel imaging modality that captures hemodynamic flow velocity and clinically relevant metrics such as vessel wall stress. These metrics are critical for assessing aneurysm rupture risk. Our model is initially trained on high resolution computational fluid dynamics (CFD) simulations and their downsampled counterparts. It is then fine-tuned on a small, harmonized dataset of paired 4D Flow MRI and CFD samples. We derive the theoretical properties of our distributional estimators and demonstrate that our framework significantly outperforms traditional deep learning approaches through real data applications. This highlights the effectiveness of distributional learning in addressing domain shift and improving super-resolution performance in clinically realistic scenarios.
Coarse-Super-Resolution-Fine Network (CoSF-Net): A Unified End-to-End Neural Network for 4D-MRI with Simultaneous Motion Estimation and Super-Resolution
Coupled-Projection Residual Network for MRI Super-Resolution
Harbin Institute of Technology(哈尔滨工业大学);Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences(中国科学院深圳先进技术研究院);Nanyang Technological University(南洋理工大学);Shenzhen University(深圳大学);Inception Institute of Artificial Intelligence(因赛普申人工智能研究院)
Synthetic Low-Field MRI Super-Resolution Via Nested U-Net Architecture
Athinoula A. Martinos Center for Biomedical Imaging(阿西努拉·A·马蒂诺斯生物医学成像中心);Massachusetts General Hospital(马萨诸塞州总医院);Harvard Medical School(哈佛医学院);Boston University(波士顿大学);Harvard University(哈佛大学)
Single-Subject Multi-View MRI Super-Resolution via Implicit Neural Representations
Department of Radiology, Weill Cornell Medicine, New York, USA(韦尔·科恩医学系放射科,纽约,美国);School of Electrical and Computer Engineering, Cornell University(康奈尔大学电气与计算机工程学院);Cornell Tech, New York, USA(康奈尔科技,纽约,美国);Weill Cornell Graduate School of Medical Sciences, New York, USA(韦尔·科恩医学科学研究生院,纽约,美国)
ShuffleUNet: Super resolution of diffusion-weighted MRIs using deep learning
Otto von Guericke University Magdeburg(马格德堡奥托·冯·格里克大学);University Hospital Magdeburg(马格德堡大学医院);German Centre for Neurodegenerative Diseases(德国神经退行性疾病中心);Center for Behavioral Brain Sciences(行为脑科学中心);Leibniz Institute for Neurobiology(莱布尼茨神经生物学研究所);MedDigit
Distillation-Driven Diffusion Model for Multi-Scale MRI Super-Resolution: Make 1.5T MRI Great Again
Massachusetts General Hospital(麻省总医院);Harvard Medical School(哈佛医学院);University Sorbonne Paris Nord(巴黎北索邦大学);Jiangsu Institute of Hematology(江苏血液学研究所);The First Affiliated Hospital of Soochow University(苏州大学附属第一医院);Henan University of Chinese Medicine(河南中医药大学);German Cancer Research Center(德国癌症研究中心);Athinoula A. Martinos Centre for Biomedical Imaging(阿西努拉·A.马蒂诺斯生物医学成像中心);Geneva University Hospital(日内瓦大学医院);IDP Institute(IDP研究所);University of Orleans(奥尔良大学)
Flexible Alignment Super-Resolution Network for Multi-Contrast MRI
Xiao Ying AI Lab(小鹰人工智能实验室);School of Electrical and Information Engineering, Tianjin University(天津大学电气自动化与信息工程学院);Peking Union Medical College Hospital(北京协和医院);School of Electronics Engineering and Computer Science, Peking University(北京大学电子工程与计算机科学学院)
Dual Arbitrary Scale Super-Resolution for Multi-Contrast MRI
Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院);The Hong Kong Polytechnic University(香港理工大学);The University of Texas at Dallas(德克萨斯大学达拉斯分校)
School of Biomedical Engineering, ShanghaiTech University(上海科技大学生物医学工程学院);Shanghai Academy of AI for Science(上海人工智能研究院);Fudan University(复旦大学);School of Computing and Communications, Lancaster University(兰卡斯特大学计算与通信学院);School of Biomedical Engineering & State Key Laboratory of Advanced Medical Materials and Devices, ShanghaiTech University(上海科技大学生物医学工程学院与先进医学材料与器械国家重点实验室);Shanghai Clinical Research and Trial Center(上海临床研究与试验中心)
Unrolled Reconstruction with Integrated Super-Resolution for Accelerated 3D LGE MRI
Kahlert School of Computing, University of Utah, Salt Lake City, UT, USA Radiology \& Imaging Sciences, University of Utah, Salt Lake City, UT, USA Cardiology, University of Utah, Salt Lake City, UT, USA
Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School(麻省总医院和哈佛医学院阿蒂努拉·A·马蒂诺斯生物医学成像中心);Department of Radiology, Massachusetts General Hospital and Harvard Medical School(麻省总医院和哈佛医学院放射科);Centre for Medical Image Computing, Department of Medical Physics and Biomedical Engineering, University College London(伦敦大学学院医学物理与生物医学工程系医学图像计算中心);Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology(麻省理工学院计算机科学与人工智能实验室);Department of Neurology and Center for Genomic Medicine, Massachusetts General Hospital and Harvard Medical School(麻省总医院和哈佛医学院神经内科与基因组医学中心);Swiss Federal Institute of Technology (ETH)(苏黎世联邦理工学院);Department of Emergency Medicine, Massachusetts General Hospital and Harvard Medical School(麻省总医院和哈佛医学院急诊医学科);Department of Neurology, Yale New Haven Hospital(耶鲁纽黑文医院神经内科);Department of Physics, Harvard University(哈佛大学物理系)
InverseSR: 3D Brain MRI Super-Resolution Using a Latent Diffusion Model
St. Francis Xavier University(圣弗朗西斯泽维尔大学);Martinos Center for Biomedical Imaging(马蒂诺生物医学成像中心);Massachusetts General Hospital(麻省总医院);Nova Scotia Health Authority(新斯科舍省卫生局);King’s College London(伦敦国王学院);University of California, Santa Cruz(加州大学圣克鲁兹分校)
Layer Selection in Feature-Based Losses Affects Image Quality and Microstructural Consistency in Deep Learning Super-Resolution of Brain Diffusion MRI
Institue for Applied Medical Informatics, University Medical Center Hamburg-Eppendorf(应用医学信息学研究所,汉堡大学医学中心);Institute of Computational Neuroscience, University Medical Center Hamburg-Eppendorf(计算神经科学研究所,汉堡大学医学中心);Center for Biomedical Artificial Intelligence (bAIome), University Medical Center Hamburg-Eppendorf(生物医学人工智能中心(bAIome),汉堡大学医学中心)
MRI Super-Resolution with Deep Learning: A Comprehensive Survey
A. I. Virtanen Institute for Molecular Sciences, Faculty of Health Sciences, University of Eastern Finland(A.I.维塔内恩分子科学研究所,健康科学学院,东芬兰大学);Harvard Medical School and Boston Children’s Hospital(哈佛医学院和波士顿儿童医院);Department of Radiology, Technical University of Munich(医学影像学系,慕尼黑技术大学);Department of Radiology, Massachusetts General Hospital(医学影像学系,麻省总医院)
Model-Guided Multi-Contrast Deep Unfolding Network for MRI Super-resolution Reconstruction
University of Science and Technology of China(中国科学技术大学);Institute of Advanced Technology(先进技术研究院);Institute of Artificial Intelligence, Hefei Comprehensive National Science Center(合肥综合性国家科学中心人工智能研究院)
4DFlowNet: Super-Resolution 4D Flow MRI using Deep Learning and Computational Fluid Dynamics
University of Auckland(奥克兰大学);University of Leeds(利兹大学);Auckland Bioengineering Institute(奥克兰生物工程研究所);Institute of Environmental Science and Research(环境科学与研究所);King's College London(伦敦国王学院)
A Diffusion-Driven Temporal Super-Resolution and Spatial Consistency Enhancement Framework for 4D MRI imaging
Paul C. Lauterbur Research Center for Biomedical Imaging(保罗·C·劳特伯生物医学成像研究中心);Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences(深圳先进技术研究院,中国科学院);University of Chinese Academy of Sciences(中国科学院大学);Pengcheng Laboratory(鹏城实验室);Faculty of Applied Sciences, Macao Polytechnic University(澳门理工学院应用科学学院)
Distributionally Robust Segmentation of Abnormal Fetal Brain 3D MRI
School of Biomedical Engineering & Imaging Sciences, King’s College London, UK(伦敦国王学院生物医学工程与成像科学学院);Department of Radiology, University Hospitals Leuven, Belgium(鲁汶大学医院放射科);Institute for Women’s Health, University College London, UK(伦敦大学学院妇女健康研究所);Department of Obstetrics and Gynaecology, University Hospitals Leuven, Belgium(鲁汶大学医院妇产科);Department of Gynecology and Obstetrics, University Hospitals Tanta, Egypt(坦塔大学医院妇产科);Department of Biomedical Imaging and Image-guided Therapy Medical University of Vienna, Austria(维也纳医科大学生物医学成像与图像引导治疗系)
深度神经网络的性能通常随着训练图像的数量而增加。然而,并非所有图像都对提高性能和鲁棒性具有相同的重要性。与非病理性病例相比,在胎儿脑部MRI中,异常会加剧发育中的脑部解剖结构的变异性。少量的异常情况,通常在用于训练的临床数据集中可用,不太可能公平地代表异常发育大脑的丰富变异性。这导致通过最大化平均性能来训练的机器学习系统偏向于非病理情况。这个问题最近被称为隐藏分层。为了适合临床使用,自动分割方法需要可靠地实现高质量的分割结果,也适用于病理病例。在本文中,我们证明了最先进的深度学习管道nnU-Net很难推广到看不见的异常情况。为了缓解这个问题,我们建议训练一个深度神经网络,以最小化数据集上每体积损失分布的百分位数。我们表明,这可以通过使用分布鲁棒优化(DRO)。DRO会自动对性能较低的训练样本进行重新加权,鼓励nnU-Net在所有情况下都能更一致地执行。我们使用368例胎儿脑T2 w MRI数据集验证了我们的方法,其中包括124例开放性脊柱裂病例的MRI和51例其他严重脑发育异常病例的MRI。