PMID- 35732795 OWN - NLM STAT- MEDLINE DCOM- 20220624 LR - 20230315 IS - 2045-2322 (Electronic) IS - 2045-2322 (Linking) VI - 12 IP - 1 DP - 2022 Jun 22 TI - Large-scale investigation of deep learning approaches for ventilated lung segmentation using multi-nuclear hyperpolarized gas MRI. PG - 10566 LID - 10.1038/s41598-022-14672-2 [doi] LID - 10566 AB - Respiratory diseases are leading causes of mortality and morbidity worldwide. Pulmonary imaging is an essential component of the diagnosis, treatment planning, monitoring, and treatment assessment of respiratory diseases. Insights into numerous pulmonary pathologies can be gleaned from functional lung MRI techniques. These include hyperpolarized gas ventilation MRI, which enables visualization and quantification of regional lung ventilation with high spatial resolution. Segmentation of the ventilated lung is required to calculate clinically relevant biomarkers. Recent research in deep learning (DL) has shown promising results for numerous segmentation problems. Here, we evaluate several 3D convolutional neural networks to segment ventilated lung regions on hyperpolarized gas MRI scans. The dataset consists of 759 helium-3 ((3)He) or xenon-129 ((129)Xe) volumetric scans and corresponding expert segmentations from 341 healthy subjects and patients with a wide range of pathologies. We evaluated segmentation performance for several DL experimental methods via overlap, distance and error metrics and compared them to conventional segmentation methods, namely, spatial fuzzy c-means (SFCM) and K-means clustering. We observed that training on combined (3)He and (129)Xe MRI scans using a 3D nn-UNet outperformed other DL methods, achieving a mean +/- SD Dice coefficient of 0.963 +/- 0.018, average boundary Hausdorff distance of 1.505 +/- 0.969 mm, Hausdorff 95th percentile of 5.754 +/- 6.621 mm and relative error of 0.075 +/- 0.039. Moreover, limited differences in performance were observed between (129)Xe and (3)He scans in the testing set. Combined training on (129)Xe and (3)He yielded statistically significant improvements over the conventional methods (p < 0.0001). In addition, we observed very strong correlation and agreement between DL and expert segmentations, with Pearson correlation of 0.99 (p < 0.0001) and Bland-Altman bias of - 0.8%. The DL approach evaluated provides accurate, robust and rapid segmentations of ventilated lung regions and successfully excludes non-lung regions such as the airways and artefacts. This approach is expected to eliminate the need for, or significantly reduce, subsequent time-consuming manual editing. CI - (c) 2022. The Author(s). FAU - Astley, Joshua R AU - Astley JR AD - Department of Oncology and Metabolism, The University of Sheffield, Sheffield, UK. AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Biancardi, Alberto M AU - Biancardi AM AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Hughes, Paul J C AU - Hughes PJC AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Marshall, Helen AU - Marshall H AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Smith, Laurie J AU - Smith LJ AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Collier, Guilhem J AU - Collier GJ AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Eaden, James A AU - Eaden JA AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Weatherley, Nicholas D AU - Weatherley ND AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. FAU - Hatton, Matthew Q AU - Hatton MQ AD - Department of Oncology and Metabolism, The University of Sheffield, Sheffield, UK. FAU - Wild, Jim M AU - Wild JM AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. AD - Insigneo Institute for In Silico Medicine, The University of Sheffield, Sheffield, UK. FAU - Tahir, Bilal A AU - Tahir BA AD - Department of Oncology and Metabolism, The University of Sheffield, Sheffield, UK. b.tahir@sheffield.ac.uk. AD - POLARIS, Department of Infection, Immunity and Cardiovascular Disease, The University of Sheffield, Sheffield, UK. b.tahir@sheffield.ac.uk. AD - Insigneo Institute for In Silico Medicine, The University of Sheffield, Sheffield, UK. b.tahir@sheffield.ac.uk. LA - eng GR - SP/14/6/31350/BHF_/British Heart Foundation/United Kingdom GR - NIHR-RP-R3-12-027/DH_/Department of Health/United Kingdom GR - MR/M008894/1/MRC_/Medical Research Council/United Kingdom PT - Journal Article PT - Research Support, Non-U.S. Gov't DEP - 20220622 PL - England TA - Sci Rep JT - Scientific reports JID - 101563288 SB - IM MH - *Deep Learning MH - Humans MH - Lung/diagnostic imaging MH - Lung Volume Measurements MH - Magnetic Resonance Imaging/methods MH - Male PMC - PMC9217976 COIS- The authors declare no competing interests. EDAT- 2022/06/23 06:00 MHDA- 2022/06/25 06:00 PMCR- 2022/06/22 CRDT- 2022/06/22 23:24 PHST- 2022/02/18 00:00 [received] PHST- 2022/06/10 00:00 [accepted] PHST- 2022/06/22 23:24 [entrez] PHST- 2022/06/23 06:00 [pubmed] PHST- 2022/06/25 06:00 [medline] PHST- 2022/06/22 00:00 [pmc-release] AID - 10.1038/s41598-022-14672-2 [pii] AID - 14672 [pii] AID - 10.1038/s41598-022-14672-2 [doi] PST - epublish SO - Sci Rep. 2022 Jun 22;12(1):10566. doi: 10.1038/s41598-022-14672-2.