Multicenter validation of synthetic FLAIR as a substitute for FLAIR sequence in acute ischemic stroke.

IF 5.8 3区医学 Q1 CLINICAL NEUROLOGY

European Stroke Journal Pub Date : 2024-08-03 DOI:10.1177/23969873241263418

Guillaume Hamon, Laurence Legrand, Ghazi Hmeydia, Guillaume Turc, Wagih Ben Hassen, Sylvain Charron, Clement Debacker, Olivier Naggara, Bertrand Thirion, Bailiang Chen, Bertrand Lapergue, Catherine Oppenheim, Joseph Benzakoun

{"title":"Multicenter validation of synthetic FLAIR as a substitute for FLAIR sequence in acute ischemic stroke.","authors":"Guillaume Hamon, Laurence Legrand, Ghazi Hmeydia, Guillaume Turc, Wagih Ben Hassen, Sylvain Charron, Clement Debacker, Olivier Naggara, Bertrand Thirion, Bailiang Chen, Bertrand Lapergue, Catherine Oppenheim, Joseph Benzakoun","doi":"10.1177/23969873241263418","DOIUrl":null,"url":null,"abstract":"Purpose: To evaluate performance of synthetic and real FLAIR for identifying early stroke in a multicenter cohort.Methods: This retrospective study was conducted using DWI and FLAIR extracted from the Endovascular Treatment in Ischemic Stroke image registry (2017-2021). The database was partitioned into subsets according to MRI field strength and manufacturer, and randomly divided into training set (70%) used for model fine-tuning, validation set (15%), and test set (15%). In test set, five readers, blinded to FLAIR sequence type, assessed DWI-FLAIR mismatch using real and synthetic FLAIR. Interobserver agreement for DWI-FLAIR rating and concordance between synthetic and real FLAIR were evaluated with kappa statistics. Sensitivity and specificity for identification of ⩽4.5 h AIS were compared in patients with known onset-to-MRI delay using McNemar's test.Results: 1454 complete MRI sets (1172 patients, median (IQR) age: 73 years (62-82); 762 women) acquired on 125 MRI units were analyzed. In test set (207 MRI), interobserver reproducibility for DWI-FLAIR mismatch labeling was substantial for real and synthetic FLAIR (Fleiss κ = 0.79 (95%CI: 0.73-0.84) and 0.77 (95%CI: 0.71-0.82), respectively). After consensus, concordance between real and synthetic FLAIR was excellent (κ = 0.85 (95%CI: 0.78-0.92)). In 141 MRI sets with known onset-to-MRI delay, diagnostic performances for ⩽4.5 h AIS identification did not differ between real and synthetic FLAIR (sensitivity: 60/71 (85%) vs 59/71 (83%), p = .56; specificity: 65/70 (93%) vs 65/70 (93%), p > 0.99).Conclusion: A deep-learning-based FLAIR fine-tuned on multicenter data can provide comparable performances to real FLAIR for early AIS identification. This approach may help reducing MR protocol duration and motion artifacts.","PeriodicalId":46821,"journal":{"name":"European Stroke Journal","volume":null,"pages":null},"PeriodicalIF":5.8000,"publicationDate":"2024-08-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"European Stroke Journal","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1177/23969873241263418","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CLINICAL NEUROLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Purpose: To evaluate performance of synthetic and real FLAIR for identifying early stroke in a multicenter cohort.

Methods: This retrospective study was conducted using DWI and FLAIR extracted from the Endovascular Treatment in Ischemic Stroke image registry (2017-2021). The database was partitioned into subsets according to MRI field strength and manufacturer, and randomly divided into training set (70%) used for model fine-tuning, validation set (15%), and test set (15%). In test set, five readers, blinded to FLAIR sequence type, assessed DWI-FLAIR mismatch using real and synthetic FLAIR. Interobserver agreement for DWI-FLAIR rating and concordance between synthetic and real FLAIR were evaluated with kappa statistics. Sensitivity and specificity for identification of ⩽4.5 h AIS were compared in patients with known onset-to-MRI delay using McNemar's test.

Results: 1454 complete MRI sets (1172 patients, median (IQR) age: 73 years (62-82); 762 women) acquired on 125 MRI units were analyzed. In test set (207 MRI), interobserver reproducibility for DWI-FLAIR mismatch labeling was substantial for real and synthetic FLAIR (Fleiss κ = 0.79 (95%CI: 0.73-0.84) and 0.77 (95%CI: 0.71-0.82), respectively). After consensus, concordance between real and synthetic FLAIR was excellent (κ = 0.85 (95%CI: 0.78-0.92)). In 141 MRI sets with known onset-to-MRI delay, diagnostic performances for ⩽4.5 h AIS identification did not differ between real and synthetic FLAIR (sensitivity: 60/71 (85%) vs 59/71 (83%), p = .56; specificity: 65/70 (93%) vs 65/70 (93%), p > 0.99).

Conclusion: A deep-learning-based FLAIR fine-tuned on multicenter data can provide comparable performances to real FLAIR for early AIS identification. This approach may help reducing MR protocol duration and motion artifacts.

查看原文本刊更多论文

合成 FLAIR 替代 FLAIR 序列用于急性缺血性中风的多中心验证。

目的：评估合成和真实 FLAIR 在多中心队列中识别早期卒中的性能：这项回顾性研究使用了从缺血性卒中血管内治疗图像登记（2017-2021 年）中提取的 DWI 和 FLAIR。根据磁共振成像场强和制造商将数据库划分为若干子集，并随机分为用于模型微调的训练集（70%）、验证集（15%）和测试集（15%）。在测试集中，五位对 FLAIR 序列类型保密的读者使用真实和合成 FLAIR 评估 DWI-FLAIR 不匹配情况。用卡帕统计法评估了 DWI-FLAIR 评级的观察者间一致性以及合成和真实 FLAIR 的一致性。使用 McNemar 检验比较了已知发病到 MRI 发生延迟的患者识别⩽4.5 h AIS 的敏感性和特异性：分析了在 125 台磁共振成像设备上获得的 1454 套完整磁共振成像（1172 名患者，中位（IQR）年龄：73 岁（62-82）；762 名女性）。在测试集（207 个 MRI）中，真实和合成 FLAIR（Fleiss κ = 0.79（95%CI：0.73-0.84）和 0.77（95%CI：0.71-0.82））DWI-FLAIR 错配标记的观察者间再现性很高。在达成共识后，真实和合成 FLAIR 的一致性非常好（κ = 0.85 (95%CI: 0.78-0.92)）。在 141 组已知发病到 MRI 延迟的 MRI 中，真实 FLAIR 和合成 FLAIR 对⩽4.5 h AIS 识别的诊断性能没有差异（灵敏度：60/71 (85%) vs 59/71 (83%)，p = .56；特异性：65/70 (93%) vs 65/70 (93%)，p > 0.99）：结论：基于深度学习的 FLAIR 经多中心数据微调后，可在早期 AIS 识别方面提供与真实 FLAIR 相当的性能。这种方法可能有助于缩短 MR 方案的持续时间并减少运动伪影。

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

European Stroke Journal Multiple-

CiteScore

7.50

自引率

6.60%

发文量

102

期刊介绍： Launched in 2016 the European Stroke Journal (ESJ) is the official journal of the European Stroke Organisation (ESO), a professional non-profit organization with over 1,400 individual members, and affiliations to numerous related national and international societies. ESJ covers clinical stroke research from all fields, including clinical trials, epidemiology, primary and secondary prevention, diagnosis, acute and post-acute management, guidelines, translation of experimental findings into clinical practice, rehabilitation, organisation of stroke care, and societal impact. It is open to authors from all relevant medical and health professions. Article types include review articles, original research, protocols, guidelines, editorials and letters to the Editor. Through ESJ, authors and researchers have gained a new platform for the rapid and professional publication of peer reviewed scientific material of the highest standards; publication in ESJ is highly competitive. The journal and its editorial team has developed excellent cooperation with sister organisations such as the World Stroke Organisation and the International Journal of Stroke, and the American Heart Organization/American Stroke Association and the journal Stroke. ESJ is fully peer-reviewed and is a member of the Committee on Publication Ethics (COPE). Issues are published 4 times a year (March, June, September and December) and articles are published OnlineFirst prior to issue publication.