深度学习实现的心脏外膜脂肪组织全自动化CT定量分析：一项多中心研究
Fully Automated CT Quantification of Epicardial Adipose Tissue by Deep Learning: A Multicenter Study.
作者信息
Commandeur Frederic, Goeller Markus, Razipour Aryabod, Cadet Sebastien, Hell Michaela M, Kwiecinski Jacek, Chen Xi, Chang Hyuk-Jae, Marwan Mohamed, Achenbach Stephan, Berman Daniel S, Slomka Piotr J, Tamarappoo Balaji K, Dey Damini
机构信息
Biomedical Imaging Research Institute (F.C., A.R., D.D.) and Department of Imaging and Medicine (S.C., J.K., X.C., D.S.B., P.J.S., B.K.T.), Cedars-Sinai Medical Center, 8700 Beverly Blvd, Taper A238, Los Angeles, CA 90048; Department of Cardiology, Friedrich-Alexander University Erlangen-Nürnberg, Erlangen, Germany (M.G., M.M.H., M.M., S.A.); and Severance Cardiovascular Hospital, Yonsei University College of Medicine, Seoul, South Korea (H.J.C.).
出版信息
Radiol Artif Intell. 2019 Nov 27;1(6):e190045. doi: 10.1148/ryai.2019190045.
PURPOSE
To evaluate the performance of deep learning for robust and fully automated quantification of epicardial adipose tissue (EAT) from multicenter cardiac CT data.
MATERIALS AND METHODS
In this multicenter study, a convolutional neural network approach was trained to quantify EAT on non-contrast material-enhanced calcium-scoring CT scans from multiple cohorts, scanners, and protocols ( = 850). Deep learning performance was compared with the performance of three expert readers and with interobserver variability in a subset of 141 scans. The deep learning algorithm was incorporated into research software. Automated EAT progression was compared with expert measurements for 70 patients with baseline and follow-up scans.
RESULTS
Automated quantification was performed in a mean (± standard deviation) time of 1.57 seconds ± 0.49, compared with 15 minutes for experts. Deep learning provided high agreement with expert manual quantification for all scans ( = 0.974; < .001), with no significant bias (0.53 cm; = .13). Manual EAT volumes measured by two experienced readers were highly correlated ( = 0.984; < .001) but with a bias of 4.35 cm ( < .001). Deep learning quantifications were highly correlated with the measurements of both experts ( = 0.973 and = 0.979; < .001), with significant bias for reader 1 (5.11 cm; < .001) but not for reader 2 (0.88 cm; = .26). EAT progression by deep learning correlated strongly with manual EAT progression ( = 0.905; < .001) in 70 patients, with no significant bias (0.64 cm; = .43), and was related to an increased noncalcified plaque burden quantified from coronary CT angiography (5.7% vs 1.8%; = .026).
CONCLUSION
Deep learning allows rapid, robust, and fully automated quantification of EAT from calcium scoring CT. It performs as well as an expert reader and can be implemented for routine cardiovascular risk assessment.© RSNA, 2019See also the commentary by Schoepf and Abadia in this issue.
目的
评估深度学习在多中心心脏CT数据中对心外膜脂肪组织（EAT）进行稳健且全自动定量分析的性能。
材料与方法
在这项多中心研究中，训练了一种卷积神经网络方法，用于在来自多个队列、扫描仪和扫描方案（n = 850）的非对比剂增强钙化积分CT扫描上对EAT进行定量分析。将深度学习性能与三位专家阅片者的性能以及141例扫描子集中的观察者间变异性进行比较。深度学习算法被整合到研究软件中。对70例有基线扫描和随访扫描的患者，将自动EAT进展情况与专家测量结果进行比较。
结果
自动定量分析的平均（±标准差）时间为1.57秒±0.49秒，而专家则需要15分钟。深度学习对所有扫描的结果与专家手动定量分析高度一致（r = 0.974；P <.001），无显著偏差（0.53 cm；P =.13）。两位经验丰富的阅片者测量的手动EAT体积高度相关（r = 0.984；P <.001），但偏差为4.35 cm（P <.001）。深度学习定量分析与两位专家的测量结果高度相关（r = 0.973和r = 0.979；P <.001），阅片者1有显著偏差（5.11 cm；P <.001），而阅片者2无显著偏差（0.88 cm；P =.26）。在70例患者中，深度学习得出的EAT进展与手动EAT进展密切相关（r = 0.905；P <.001），无显著偏差（0.64 cm；P =.43），并且与冠状动脉CT血管造影定量分析得出的非钙化斑块负荷增加相关（5.7%对1.8%；P =.026）。
结论
深度学习能够从钙化积分CT中快速、稳健且全自动地对EAT进行定量分析。其性能与专家阅片者相当，可用于常规心血管风险评估。© RSNA，2019另见本期Schoepf和Abadia的评论。
Zotero 插件