Deep learning has innovated the field of computational imaging. One of its bottlenecks is unavailable or insufficient training data. This article reviews an emerging paradigm, imaging physics-based data synthesis (IPADS), that can provide huge training data in biomedical magnetic resonance without or with few real data. Following the physical law of magnetic resonance, IPADS generates signals from differential equations or analytical solution models, making the learning more scalable, explainable, and better protecting privacy. Key components of IPADS learning, including signal generation models, basic deep learning network structures, enhanced data generation, and learning methods are discussed. Great potentials of IPADS have been demonstrated by representative applications in fast imaging, ultrafast signal reconstruction and accurate parameter quantification. Finally, open questions and future work have been discussed.
Forecasting-based Biomedical Time-series Data Synthesis for Open Data and Robust AI
School of Electrical Engineering, KAIST, Daejeon, Republic of Korea(韩国成均馆大学电气工程学院);Department of Applied AI, Hansung University, Seoul, Republic of Korea(汉城市立大学应用人工智能系)
A Reinforcement Learning Approach to Synthetic Data Generation
Department of Computer Science, Princeton University(普林斯顿大学计算机科学系);Department of Biomedical Informatics, Vanderbilt University Medical Center(范德比尔特大学医学中心生物医学信息学系);Department of Biostatistics, Vanderbilt University Medical Center(范德比尔特大学医学中心生物统计学系);Department of Computer Science, Vanderbilt University(范德比尔特大学计算机科学系);John F. Hardesty Department of Ophthalmology and Visual Sciences, Washington University(华盛顿大学约翰·F·哈里斯德眼科学与视觉科学系)
SYNTA: A novel approach for deep learning-based image analysis in muscle histopathology using photo-realistic synthetic data
MIRA Vision Microscopy GmbH(MIRA Vision 显微镜有限公司);Friedrich-Alexander University Erlangen-Nürnberg (FAU)(埃尔朗根-纽伦堡弗里德里希-亚历山大大学);Universitätsklinikum Erlangen(埃尔朗根大学医院);German Aerospace Center (DLR)(德国航空航天中心);University of Cologne(科隆大学);Leibniz-Institut für Analytische Wissenschaften - ISAS - e.V.(莱布尼茨分析科学研究所)
本文提出 SYNTA 方法,通过生成照片级逼真的合成生物医学图像作为深度学习训练数据,在无需人工标注的情况下实现了对真实肌肉组织切片的专家级分割,为 GAN 提供了可解释且可控的替代方案。
Privacy-Preserving Statistical Data Generation: Application to Sepsis Detection
Instituto de Ingeniería del Conocimiento(知识工程研究所);Hospital Universitario Son Llátzer(松利亚茨大学医院);Fundación Instituto de Investigación Sanitaria Illes Balears(巴利阿里群岛卫生研究所基金会);BBforTAI;Universidad Autónoma de Madrid(马德里自治大学)
CytoDiff: AI-Driven Cytomorphology Image Synthesis for Medical Diagnostics
Institute of AI for Health, Helmholtz Zentrum München - German Research Center for Environmental Health(人工智能与健康研究所,慕尼黑德国环境健康研究中心);Department of Medicine III, Ludwig-Maximilian-University Hospital(第三医学部,路德维希-马克西米利安大学医院);DKTK, German Cancer Consortium(德国癌症联盟,DKTK);Escola Superior de Comerç Internacional, Universitat Pompeu Fabra (ESCI-UPF)(国际商业学校,庞培法布拉大学(ESCI-UPF))
Enhanced Magnetic Resonance Image Synthesis with Contrast-Aware Generative Adversarial Networks
Friedrich-Alexander Universität Erlangen-Nürnberg(弗里德里希-亚历山大大学埃尔兰根-纽伦堡分校);Technical University of Applied Sciences Amberg-Weiden(安贝格-魏登应用技术大学);Siemens Healthcare GmbH(西门子医疗有限公司)
Convex space learning for tabular synthetic data generation
Institute of Computer Science, University of Rostock(罗斯托克大学计算机科学学院);Stellenbosch Institute for Advanced Study(斯泰伦博斯高等研究院);School of Data Science, Indian Institute of Science Education and Research Thiruvananthapuram(印度 Thiruvananthapuram 科学教育与研究学院数据科学学院)
Magnetic Resonance Spectroscopy Quantification using Deep Learning
Univ. Lyon(里昂大学);INSA-Lyon(里昂国家应用科学学院);Université Claude Bernard Lyon 1(里昂第一大学克洛德·贝尔纳大学);UJM-Saint Etienne(圣艾蒂安大学);CNRS(法国国家科学研究中心);Inserm(法国国家健康与医学研究院)