Pulmonary 129Xe MRI: CNN Registration and Segmentation to Generate Ventilation Defect Percent with Multi-center Validation

Ali Mozaffaripour; Alexander M Matheson; Omar Rahman; Maksym Sharma; Harkiran K Kooner; Marrissa J McIntosh; Jonathan Rayment; Rachel L Eddy; Sarah Svenningsen; Grace Parraga

doi:10.1016/j.acra.2024.10.029

Pulmonary ¹²⁹Xe MRI: CNN Registration and Segmentation to Generate Ventilation Defect Percent with Multi-center Validation

Acad Radiol. 2024 Nov 23:S1076-6332(24)00789-X. doi: 10.1016/j.acra.2024.10.029. Online ahead of print.

Authors

Affiliations

¹ Robarts Research Institute, Western University, London, Canada; School of Biomedical Engineering, Western University, London, Canada.
² Robarts Research Institute, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada; Cincinnati Children's Hospital, Cincinnati, Ohio, USA.
³ Robarts Research Institute, Western University, London, Canada.
⁴ Robarts Research Institute, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada; Department of Radiation Oncology, University of Pennsylvania, Philadelphia, Pennsylvania, USA.
⁵ Robarts Research Institute, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada; Department of Radiation Oncology, Johns Hopkins University, Baltimore, Maryland, USA.
⁶ Robarts Research Institute, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada; Department of Radiology, University of Iowa, Iowa City, Iowa, USA.
⁷ BC Children's Hospital Research Institute, Vancouver, Canada.
⁸ Centre for Heart Lung Innovation, St. Paul's Hospital, Vancouver, Canada.
⁹ Division of Respirology, Department of Medicine, McMaster University, Hamilton, Canada.
¹⁰ Robarts Research Institute, Western University, London, Canada; School of Biomedical Engineering, Western University, London, Canada; Department of Medical Biophysics, Western University, London, Canada; Division of Respirology, Department of Medicine, Western University, London, Canada. Electronic address: [email protected].

PMID: 39581785
DOI: 10.1016/j.acra.2024.10.029

Abstract

Rationale and objectives: Hyperpolarized ¹²⁹Xe MRI quantifies ventilation-defect-percent (VDP), the ratio of ¹²⁹Xe signal-void to the anatomic ¹H 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 ¹²⁹Xe MRI VDP.

Materials and methods: ¹H and ¹²⁹Xe 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.

Results: 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 ¹H and ¹²⁹Xe registrations agreed (MAE=0.3°, R² =0.95 rotation; 1.1%, R² =0.79 scaling; 0.2/0.5px, R² =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 R²/ρ = .97/.95, bias=-0.5%; Site2 R²/ρ = .85/.93, bias=-0.9%; Site3 R²/ρ = .95/.89, bias=-0.8%, all p < .001).

Conclusion: 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.

Keywords: Convolutional neural network; External validation; Functional lung imaging; Registration; Segmentation.