Hasenstab Kyle A, Tabalon Joseph, Yuan Nancy, Retson Tara, Hsiao Albert
Department of Radiology, University of California San Diego, 9500 Gilman Dr, San Diego, CA 92093 (K.A.H., N.Y., T.R., A.H.); and Department of Mathematics and Statistics, San Diego State University, San Diego, Calif (K.A.H., J.T.).
Radiol Artif Intell. 2021 Nov 10;4(1):e210211. doi: 10.1148/ryai.2021210211. eCollection 2022 Jan.
To develop a convolutional neural network (CNN)-based deformable lung registration algorithm to reduce computation time and assess its potential for lobar air trapping quantification.
In this retrospective study, a CNN algorithm was developed to perform deformable registration of lung CT (LungReg) using data on 9118 patients from the COPDGene Study (data collected between 2007 and 2012). Loss function constraints included cross-correlation, displacement field regularization, lobar segmentation overlap, and the Jacobian determinant. LungReg was compared with a standard diffeomorphic registration (SyN) for lobar Dice overlap, percentage voxels with nonpositive Jacobian determinants, and inference runtime using paired tests. Landmark colocalization error (LCE) across 10 patients was compared using a random effects model. Agreement between LungReg and SyN air trapping measurements was assessed using intraclass correlation coefficient. The ability of LungReg versus SyN emphysema and air trapping measurements to predict Global Initiative for Chronic Obstructive Lung Disease (GOLD) stages was compared using area under the receiver operating characteristic curves.
Average performance of LungReg versus SyN showed lobar Dice overlap score of 0.91-0.97 versus 0.89-0.95, respectively ( < .001); percentage voxels with nonpositive Jacobian determinant of 0.04 versus 0.10, respectively ( < .001); inference run time of 0.99 second (graphics processing unit) and 2.27 seconds (central processing unit) versus 418.46 seconds (central processing unit) ( < .001); and LCE of 7.21 mm versus 6.93 mm ( < .001). LungReg and SyN whole-lung and lobar air trapping measurements achieved excellent agreement (intraclass correlation coefficients > 0.98). LungReg versus SyN area under the receiver operating characteristic curves for predicting GOLD stage were not statistically different (range, 0.88-0.95 vs 0.88-0.95, respectively; = .31-.95).
CNN-based deformable lung registration is accurate and fully automated, with runtime feasible for clinical lobar air trapping quantification, and has potential to improve diagnosis of small airway diseases. Air Trapping, Convolutional Neural Network, Deformable Registration, Small Airway Disease, CT, Lung, Semisupervised Learning, Unsupervised Learning © RSNA, 2021
开发一种基于卷积神经网络（CNN）的可变形肺配准算法，以减少计算时间，并评估其在叶间气体陷闭定量分析方面的潜力。
在这项回顾性研究中，利用慢性阻塞性肺疾病基因研究（COPDGene研究，2007年至2012年收集的数据）中9118例患者的数据，开发了一种CNN算法来进行肺CT的可变形配准（LungReg）。损失函数约束包括互相关、位移场正则化、叶间分割重叠和雅可比行列式。使用配对检验，比较LungReg与标准的微分同胚配准（SyN）在叶间Dice重叠、具有非正雅可比行列式的体素百分比以及推理运行时间方面的差异。使用随机效应模型比较10例患者的地标共定位误差（LCE）。使用组内相关系数评估LungReg与SyN气体陷闭测量值之间的一致性。使用受试者操作特征曲线下面积比较LungReg与SyN在肺气肿和气体陷闭测量方面预测慢性阻塞性肺疾病全球倡议（GOLD）分期的能力。
LungReg与SyN的平均性能比较显示，叶间Dice重叠分数分别为0.91 - 0.97和0.89 - 0.95（P <.001）；具有非正雅可比行列式的体素百分比分别为0.04和0.10（P <.001）；推理运行时间为0.99秒（图形处理单元）和2.27秒（中央处理器），而SyN为418.46秒（中央处理器）（P <.001）；LCE分别为7.21毫米和6.93毫米（P <.001）。LungReg与SyN全肺和叶间气体陷闭测量值具有极好的一致性（组内相关系数> 0.98）。LungReg与SyN在预测GOLD分期的受试者操作特征曲线下面积无统计学差异（范围分别为0.88 - 0.95和0.88 - 0.95；P = 0.31 - 0.95）。
基于CNN的可变形肺配准准确且完全自动化，运行时间对于临床叶间气体陷闭定量分析是可行的，并且有潜力改善小气道疾病的诊断。气体陷闭、卷积神经网络、可变形配准、小气道疾病、CT、肺、半监督学习、无监督学习 © RSNA，2021
