Deep-learning-based image reconstruction with limited data: generating synthetic raw data using deep learning.

IF 2 4区 医学 Q3 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING
Frank Zijlstra, Peter Thomas While
{"title":"Deep-learning-based image reconstruction with limited data: generating synthetic raw data using deep learning.","authors":"Frank Zijlstra, Peter Thomas While","doi":"10.1007/s10334-024-01193-4","DOIUrl":null,"url":null,"abstract":"<p><strong>Object: </strong>Deep learning has shown great promise for fast reconstruction of accelerated MRI acquisitions by learning from large amounts of raw data. However, raw data is not always available in sufficient quantities. This study investigates synthetic data generation to complement small datasets and improve reconstruction quality.</p><p><strong>Materials and methods: </strong>An adversarial auto-encoder was trained to generate phase and coil sensitivity maps from magnitude images, which were combined into synthetic raw data. On a fourfold accelerated MR reconstruction task, deep-learning-based reconstruction networks were trained with varying amounts of training data (20 to 160 scans). Test set performance was compared between baseline experiments and experiments that incorporated synthetic training data.</p><p><strong>Results: </strong>Training with synthetic raw data showed decreasing reconstruction errors with increasing amounts of training data, but importantly this was magnitude-only data, rather than real raw data. For small training sets, training with synthetic data decreased the mean absolute error (MAE) by up to 7.5%, whereas for larger training sets the MAE increased by up to 2.6%.</p><p><strong>Discussion: </strong>Synthetic raw data generation improved reconstruction quality in scenarios with limited training data. A major advantage of synthetic data generation is that it allows for the reuse of magnitude-only datasets, which are more readily available than raw datasets.</p>","PeriodicalId":18067,"journal":{"name":"Magnetic Resonance Materials in Physics, Biology and Medicine","volume":" ","pages":"1059-1076"},"PeriodicalIF":2.0000,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11582256/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Magnetic Resonance Materials in Physics, Biology and Medicine","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1007/s10334-024-01193-4","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2024/8/29 0:00:00","PubModel":"Epub","JCR":"Q3","JCRName":"RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING","Score":null,"Total":0}
引用次数: 0

Abstract

Object: Deep learning has shown great promise for fast reconstruction of accelerated MRI acquisitions by learning from large amounts of raw data. However, raw data is not always available in sufficient quantities. This study investigates synthetic data generation to complement small datasets and improve reconstruction quality.

Materials and methods: An adversarial auto-encoder was trained to generate phase and coil sensitivity maps from magnitude images, which were combined into synthetic raw data. On a fourfold accelerated MR reconstruction task, deep-learning-based reconstruction networks were trained with varying amounts of training data (20 to 160 scans). Test set performance was compared between baseline experiments and experiments that incorporated synthetic training data.

Results: Training with synthetic raw data showed decreasing reconstruction errors with increasing amounts of training data, but importantly this was magnitude-only data, rather than real raw data. For small training sets, training with synthetic data decreased the mean absolute error (MAE) by up to 7.5%, whereas for larger training sets the MAE increased by up to 2.6%.

Discussion: Synthetic raw data generation improved reconstruction quality in scenarios with limited training data. A major advantage of synthetic data generation is that it allows for the reuse of magnitude-only datasets, which are more readily available than raw datasets.

Abstract Image

基于深度学习的有限数据图像重建:利用深度学习生成合成原始数据。
目标:通过对大量原始数据进行学习,深度学习在加速磁共振成像采集的快速重建方面大有可为。然而,原始数据的数量并不总是充足。本研究调查了合成数据生成,以补充小型数据集并提高重建质量:对对抗性自动编码器进行了训练,以从幅值图像生成相位和线圈灵敏度图,并将其合并为合成原始数据。在一项四倍加速磁共振重建任务中,基于深度学习的重建网络在不同数量的训练数据(20 到 160 次扫描)下进行了训练。比较了基线实验和包含合成训练数据的实验的测试集性能:结果:使用合成原始数据进行的训练显示,随着训练数据量的增加,重建误差也在减少,但重要的是,这只是幅度数据,而不是真实的原始数据。对于较小的训练集,使用合成数据进行训练可使平均绝对误差(MAE)降低 7.5%,而对于较大的训练集,平均绝对误差可增加 2.6%:讨论:合成原始数据的生成提高了训练数据有限情况下的重建质量。合成数据生成的一个主要优势是,它允许重复使用仅震级数据集,这些数据集比原始数据集更容易获得。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
CiteScore
4.60
自引率
0.00%
发文量
58
审稿时长
>12 weeks
期刊介绍: MAGMA is a multidisciplinary international journal devoted to the publication of articles on all aspects of magnetic resonance techniques and their applications in medicine and biology. MAGMA currently publishes research papers, reviews, letters to the editor, and commentaries, six times a year. The subject areas covered by MAGMA include: advances in materials, hardware and software in magnetic resonance technology, new developments and results in research and practical applications of magnetic resonance imaging and spectroscopy related to biology and medicine, study of animal models and intact cells using magnetic resonance, reports of clinical trials on humans and clinical validation of magnetic resonance protocols.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信