Mozaffaripour Ali, Matheson Alexander M, Rahman Omar, Sharma Maksym, Kooner Harkiran K, McIntosh Marrissa J, Rayment Jonathan, Eddy Rachel L, Svenningsen Sarah, Parraga Grace
Robarts Research Institute, Western University, London, Canada; School of Biomedical Engineering, Western University, London, Canada.
Robarts Research Institute, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada; Cincinnati Children's Hospital, Cincinnati, Ohio, USA.
Acad Radiol. 2025 Mar;32(3):1734-1742. doi: 10.1016/j.acra.2024.10.029. Epub 2024 Nov 23.
Hyperpolarized Xe MRI quantifies ventilation-defect-percent (VDP), the ratio of Xe signal-void to the anatomic H MRI thoracic-cavity-volume. VDP is associated with airway inflammation and disease control and serves as a treatable trait in therapy studies. Semi-automated VDP pipelines require time-intensive observer interactions. Current convolutional neural network (CNN) approaches for quantifying VDP lack external validation, which limits multicenter utilization. Our objective was to develop an automated and externally validated deep-learning pipeline to quantify pulmonary Xe MRI VDP.
H and Xe MRI data from the primary site (Site1) were used to train and test a CNN segmentation and registration pipeline, while two independent sites (Site2 and Site3) provided external validation. Semi-automated and CNN-based registration error was measured using mean-absolute-error (MAE) while segmentation error was measured using generalized-Dice-similarity coefficient (gDSC). CNN and semi-automated VDP were compared using linear regression and Bland-Altman analysis.
Training/testing used data from 205 participants (healthy volunteers, asthma, COPD, long-COVID; mean age=54 ± 16y; 119 females) from Site1. External validation used data from 71 participants. CNN and semi-automated H and Xe registrations agreed (MAE=0.3°, R =0.95 rotation; 1.1%, R =0.79 scaling; 0.2/0.5px, R =0.96/0.95, x/y-translation; all p < .001). Thoracic-cavity and ventilation segmentations were also spatially corresponding (gDSC=0.92 and 0.88, respectively). CNN VDP correlated with semi-automated VDP (Site1 R/ρ = .97/.95, bias=-0.5%; Site2 R/ρ = .85/.93, bias=-0.9%; Site3 R/ρ = .95/.89, bias=-0.8%, all p < .001).
An externally validated CNN registration/segmentation model demonstrated strong agreement with low error compared to the semi-automated method. CNN and semi-automated registrations, thoracic-cavity-volume and ventilation-volume segmentations were highly correlated with high gDSC for the datasets.
超极化氙气磁共振成像（MRI）可量化通气缺陷百分比（VDP），即氙气信号缺失与解剖学氢质子MRI胸腔容积的比值。VDP与气道炎症和疾病控制相关，并且在治疗研究中是一个可治疗的特征。半自动VDP流程需要耗费大量时间的观察者交互。目前用于量化VDP的卷积神经网络（CNN）方法缺乏外部验证，这限制了其在多中心的应用。我们的目的是开发一种自动化且经过外部验证的深度学习流程，以量化肺部氙气MRI的VDP。
来自主要站点（站点1）的氢质子和氙气MRI数据用于训练和测试CNN分割与配准流程，而两个独立站点（站点2和站点3）提供外部验证。使用平均绝对误差（MAE）测量半自动和基于CNN的配准误差，同时使用广义骰子相似系数（gDSC）测量分割误差。使用线性回归和布兰德-奥特曼分析比较CNN和半自动VDP。
训练/测试使用了来自站点1的205名参与者（健康志愿者、哮喘患者、慢性阻塞性肺疾病患者、新冠后患者；平均年龄=54±16岁；119名女性）的数据。外部验证使用了71名参与者的数据。CNN和半自动氢质子及氙气配准结果一致（MAE=0.3°，旋转R=0.95；缩放1.1%，R=0.79；x/y平移0.2/0.5像素，R=0.96/0.95；所有p<0.001）。胸腔和通气分割在空间上也相互对应（gDSC分别为0.92和0.88）。CNN VDP与半自动VDP相关（站点1 R/ρ=0.97/0.95，偏差=-0.5%；站点2 R/ρ=0.85/0.93，偏差=-0.9%；站点3 R/ρ=0.95/0.89，偏差=-0.8%，所有p<0.001）。
与半自动方法相比，经过外部验证的CNN配准/分割模型显示出高度一致性且误差较低。对于数据集，CNN和半自动配准、胸腔容积和通气容积分割具有高度相关性且gDSC较高。
