用于高帧频心脏显像 MRI 的变形编码深度学习变换器

IF 3.8 Q1 RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING Radiology. Cardiothoracic imaging Pub Date : 2024-06-01 DOI:10.1148/ryct.230177
Manuel A Morales, Fahime Ghanbari, Shiro Nakamori, Salah Assana, Amine Amyar, Siyeop Yoon, Jennifer Rodriguez, Martin S Maron, Ethan J Rowin, Jiwon Kim, Robert M Judd, Jonathan W Weinsaft, Reza Nezafat
{"title":"用于高帧频心脏显像 MRI 的变形编码深度学习变换器","authors":"Manuel A Morales, Fahime Ghanbari, Shiro Nakamori, Salah Assana, Amine Amyar, Siyeop Yoon, Jennifer Rodriguez, Martin S Maron, Ethan J Rowin, Jiwon Kim, Robert M Judd, Jonathan W Weinsaft, Reza Nezafat","doi":"10.1148/ryct.230177","DOIUrl":null,"url":null,"abstract":"<p><p>Purpose To develop a deep learning model for increasing cardiac cine frame rate while maintaining spatial resolution and scan time. Materials and Methods A transformer-based model was trained and tested on a retrospective sample of cine images from 5840 patients (mean age, 55 years ± 19 [SD]; 3527 male patients) referred for clinical cardiac MRI from 2003 to 2021 at nine centers; images were acquired using 1.5- and 3-T scanners from three vendors. Data from three centers were used for training and testing (4:1 ratio). The remaining data were used for external testing. Cines with downsampled frame rates were restored using linear, bicubic, and model-based interpolation. The root mean square error between interpolated and original cine images was modeled using ordinary least squares regression. In a prospective study of 49 participants referred for clinical cardiac MRI (mean age, 56 years ± 13; 25 male participants) and 12 healthy participants (mean age, 51 years ± 16; eight male participants), the model was applied to cines acquired at 25 frames per second (fps), thereby doubling the frame rate, and these interpolated cines were compared with actual 50-fps cines. The preference of two readers based on perceived temporal smoothness and image quality was evaluated using a noninferiority margin of 10%. Results The model generated artifact-free interpolated images. Ordinary least squares regression analysis accounting for vendor and field strength showed lower error (<i>P</i> < .001) with model-based interpolation compared with linear and bicubic interpolation in internal and external test sets. The highest proportion of reader choices was \"no preference\" (84 of 122) between actual and interpolated 50-fps cines. The 90% CI for the difference between reader proportions favoring collected (15 of 122) and interpolated (23 of 122) high-frame-rate cines was -0.01 to 0.14, indicating noninferiority. Conclusion A transformer-based deep learning model increased cardiac cine frame rates while preserving both spatial resolution and scan time, resulting in images with quality comparable to that of images obtained at actual high frame rates. <b>Keywords:</b> Functional MRI, Heart, Cardiac, Deep Learning, High Frame Rate <i>Supplemental material is available for this article.</i> © RSNA, 2024.</p>","PeriodicalId":21168,"journal":{"name":"Radiology. Cardiothoracic imaging","volume":null,"pages":null},"PeriodicalIF":3.8000,"publicationDate":"2024-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11211941/pdf/","citationCount":"0","resultStr":"{\"title\":\"Deformation-encoding Deep Learning Transformer for High-Frame-Rate Cardiac Cine MRI.\",\"authors\":\"Manuel A Morales, Fahime Ghanbari, Shiro Nakamori, Salah Assana, Amine Amyar, Siyeop Yoon, Jennifer Rodriguez, Martin S Maron, Ethan J Rowin, Jiwon Kim, Robert M Judd, Jonathan W Weinsaft, Reza Nezafat\",\"doi\":\"10.1148/ryct.230177\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Purpose To develop a deep learning model for increasing cardiac cine frame rate while maintaining spatial resolution and scan time. Materials and Methods A transformer-based model was trained and tested on a retrospective sample of cine images from 5840 patients (mean age, 55 years ± 19 [SD]; 3527 male patients) referred for clinical cardiac MRI from 2003 to 2021 at nine centers; images were acquired using 1.5- and 3-T scanners from three vendors. Data from three centers were used for training and testing (4:1 ratio). The remaining data were used for external testing. Cines with downsampled frame rates were restored using linear, bicubic, and model-based interpolation. The root mean square error between interpolated and original cine images was modeled using ordinary least squares regression. In a prospective study of 49 participants referred for clinical cardiac MRI (mean age, 56 years ± 13; 25 male participants) and 12 healthy participants (mean age, 51 years ± 16; eight male participants), the model was applied to cines acquired at 25 frames per second (fps), thereby doubling the frame rate, and these interpolated cines were compared with actual 50-fps cines. The preference of two readers based on perceived temporal smoothness and image quality was evaluated using a noninferiority margin of 10%. Results The model generated artifact-free interpolated images. Ordinary least squares regression analysis accounting for vendor and field strength showed lower error (<i>P</i> < .001) with model-based interpolation compared with linear and bicubic interpolation in internal and external test sets. The highest proportion of reader choices was \\\"no preference\\\" (84 of 122) between actual and interpolated 50-fps cines. The 90% CI for the difference between reader proportions favoring collected (15 of 122) and interpolated (23 of 122) high-frame-rate cines was -0.01 to 0.14, indicating noninferiority. Conclusion A transformer-based deep learning model increased cardiac cine frame rates while preserving both spatial resolution and scan time, resulting in images with quality comparable to that of images obtained at actual high frame rates. <b>Keywords:</b> Functional MRI, Heart, Cardiac, Deep Learning, High Frame Rate <i>Supplemental material is available for this article.</i> © RSNA, 2024.</p>\",\"PeriodicalId\":21168,\"journal\":{\"name\":\"Radiology. Cardiothoracic imaging\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":3.8000,\"publicationDate\":\"2024-06-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11211941/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Radiology. Cardiothoracic imaging\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1148/ryct.230177\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Radiology. Cardiothoracic imaging","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1148/ryct.230177","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"RADIOLOGY, NUCLEAR MEDICINE & MEDICAL IMAGING","Score":null,"Total":0}
引用次数: 0

摘要

目的 开发一种深度学习模型,在保持空间分辨率和扫描时间的同时提高心脏显像帧频。材料与方法 在九个中心 2003 年至 2021 年期间转诊的 5840 名临床心脏 MRI 患者(平均年龄 55 岁 ± 19 [SD];3527 名男性患者)的回顾性电影图像样本上训练和测试了一个基于变压器的模型;图像是使用三个供应商的 1.5-T 和 3-T 扫描仪采集的。三个中心的数据用于培训和测试(4:1 比例)。其余数据用于外部测试。使用线性插值、双三次插值和基于模型的插值还原了降采样帧率的正片。插值图像与原始电影图像之间的均方根误差采用普通最小二乘法回归建模。在一项针对 49 名转诊至临床心脏磁共振成像的患者(平均年龄 56 岁 ± 13 岁;25 名男性患者)和 12 名健康患者(平均年龄 51 岁 ± 16 岁;8 名男性患者)进行的前瞻性研究中,该模型被应用于以每秒 25 帧(fps)获取的动态影像,从而将帧频提高了一倍,并将这些插值动态影像与实际的 50 帧/秒动态影像进行了比较。两位读者根据所感知的时间平滑度和图像质量的偏好进行了评估,非劣效差为 10%。结果 该模型生成了无伪影的插值图像。考虑到供应商和场强的普通最小二乘法回归分析表明,在内部和外部测试集中,基于模型的插值与线性插值和双三次插值相比,误差更小(P < .001)。在读者的选择中,"无偏好 "的比例最高(122 人中有 84 人)。读者对收集的(122 人中的 15 人)和插值的(122 人中的 23 人)高帧率电影的偏好比例差异的 90% CI 为-0.01 到 0.14,表明两者之间不存在劣势。结论 基于变压器的深度学习模型在保持空间分辨率和扫描时间的同时提高了心脏超声帧率,从而获得了与实际高帧率下获得的图像质量相当的图像。关键词功能磁共振成像 心脏 深度学习 高帧率 本文有补充材料。© RSNA, 2024.
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
Deformation-encoding Deep Learning Transformer for High-Frame-Rate Cardiac Cine MRI.

Purpose To develop a deep learning model for increasing cardiac cine frame rate while maintaining spatial resolution and scan time. Materials and Methods A transformer-based model was trained and tested on a retrospective sample of cine images from 5840 patients (mean age, 55 years ± 19 [SD]; 3527 male patients) referred for clinical cardiac MRI from 2003 to 2021 at nine centers; images were acquired using 1.5- and 3-T scanners from three vendors. Data from three centers were used for training and testing (4:1 ratio). The remaining data were used for external testing. Cines with downsampled frame rates were restored using linear, bicubic, and model-based interpolation. The root mean square error between interpolated and original cine images was modeled using ordinary least squares regression. In a prospective study of 49 participants referred for clinical cardiac MRI (mean age, 56 years ± 13; 25 male participants) and 12 healthy participants (mean age, 51 years ± 16; eight male participants), the model was applied to cines acquired at 25 frames per second (fps), thereby doubling the frame rate, and these interpolated cines were compared with actual 50-fps cines. The preference of two readers based on perceived temporal smoothness and image quality was evaluated using a noninferiority margin of 10%. Results The model generated artifact-free interpolated images. Ordinary least squares regression analysis accounting for vendor and field strength showed lower error (P < .001) with model-based interpolation compared with linear and bicubic interpolation in internal and external test sets. The highest proportion of reader choices was "no preference" (84 of 122) between actual and interpolated 50-fps cines. The 90% CI for the difference between reader proportions favoring collected (15 of 122) and interpolated (23 of 122) high-frame-rate cines was -0.01 to 0.14, indicating noninferiority. Conclusion A transformer-based deep learning model increased cardiac cine frame rates while preserving both spatial resolution and scan time, resulting in images with quality comparable to that of images obtained at actual high frame rates. Keywords: Functional MRI, Heart, Cardiac, Deep Learning, High Frame Rate Supplemental material is available for this article. © RSNA, 2024.

求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
CiteScore
20.40
自引率
1.40%
发文量
0
期刊介绍:
期刊最新文献
Association of Aortic Stiffness with Heart Failure with Preserved Ejection Fraction: New Insights into the Ventriculoarterial Connection. Severe Quadricuspid Pulmonary Valve Stenosis Presenting as Scrotal Edema. Collateral Circulation in Coarctation. Geometric Changes in Mitral Valve Apparatus during Long-term Cardiac Resynchronization Therapy as Assessed with Cardiac CT. Clinically Relevant Extracardiac Findings at Cardiac Imaging: Insights from the European MR/CT Registry.
×
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
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 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 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1