The impact of segmentation on whole‐lung functional MRI quantification: Repeatability and reproducibility from multiple human observers and an artificial neural network

Abstract

Purpose To investigate the repeatability and reproducibility of lung segmentation and their impact on the quantitative outcomes from functional pulmonary MRI. Additionally, to validate an artificial neural network (ANN) to accelerate whole‐lung quantification. Method Ten healthy children and 25 children with cystic fibrosis underwent matrix pencil decomposition MRI (MP‐MRI). Impaired relative fractional ventilation (RFV) and relative perfusion (RQ) from MP‐MRI were compared using whole‐lung segmentation performed by a physician at two time‐points (At1 and At2), by an MRI technician (B), and by an ANN (C). Repeatability and reproducibility were assess with Dice similarity coefficient (DSC), paired t‐test and Intraclass‐correlation coefficient (ICC). Results The repeatability within an observer (At1 vs At2) resulted in a DSC of 0.94 ± 0.01 (mean ± SD) and an unsystematic difference of −0.01% for RFV (P = .92) and +0.1% for RQ (P = .21). The reproducibility between human observers (At1 vs B) resulted in a DSC of 0.88 ± 0.02, and a systematic absolute difference of −0.81% (P < .001) for RFV and −0.38% (P = .037) for RQ. The reproducibility between human and the ANN (At1 vs C) resulted in a DSC of 0.89 ± 0.03 and a systematic absolute difference of −0.36% for RFV (P = .017) and −0.35% for RQ (P = .002). The ICC was >0.98 for all variables and comparisons. Conclusions Despite high overall agreement, there were systematic differences in lung segmentation between observers. This needs to be considered for longitudinal studies and could be overcome by using an ANN, which performs as good as human observers and fully automatizes MP‐MRI post‐processing.

Deep learning-based segmentation of the lung in MR-images acquired by a stack-of-spirals trajectory at ultra-short echo-times

“…As can be seen in Figs. 1 , 2 and 3 , we did not implement a separate step to eliminate vessels from the lung labels as performed in [ 27 ]. That might impair results for volume and ventilation even if the segmentations show high similarity to the manual post processing.…”

“…With the growing success of exploiting machine learning in general, a plethora of post-processing techniques based on artificial neural networks (ANN) has also been proposed for medical imaging lately. Since, empirically, the human eye seems to be able to discriminate the lung parenchyma from other tissues regardless of inhomogeneities or image artifacts quite well, and ANNs are particularly well suited for perceptual tasks, corresponding methods have lately been implemented and tested also for lung imaging with promising results [ 22 – 27 ].…”

Effect of Salbutamol on Lung Ventilation in Children with Cystic Fibrosis: Comprehensive Assessment Using Spirometry, Multiple-Breath Washout, and Functional Lung Magnetic Resonance Imaging

“…Multiple coronal slices were acquired to cover the whole lung, in a posterior to anterior order. From the baseline images, the lung parenchyma was segmented as previously described [22]. Then the regional fractional lung ventilation and relative perfusion were assessed using a matrix-pencil algorithm [23].…”

Defect distribution index: A novel metric for functional lung MRI in cystic fibrosis

“…Subsequently, fractional ventilation and relative perfusion maps were obtained from the registered time-series using the MP method (Figure 1A,B). 19 Automatic segmentation with vessel exclusion was performed on the baseline images using an artificial neural network as previously described, 28 after which the DP was determined. 20,29 The DP equals the number of defect lung voxels divided by the number of lung area voxels for each slice.…”