A Segmentation Pipeline for Measuring Pulmonary Ventilation Suitable for Clinical Workflows and Decision-making
Fumin Guo1, Khadija Sheikh1, Rachel Eddy1, Dante PI Capaldi1, David G McCormack2, Aaron Fenster1, and Grace Parraga1

1Robarts Research Institute, The University of Western Ontario, London, ON, Canada, 2Department of Medicine, The University of Western Ontario, London, ON, Canada

Synopsis

Clinical translation of hyperpolarized 129Xe MRI for large-scale and multi-centre applications requires image analysis tools that can provide clinically-acceptable measurements of pulmonary information. Here we proposed a pipeline that consists of 1H-129Xe registration, segmentation and ventilation defects generation for regional and quantitative evaluation of 129Xe ventilation. 1H-129Xe registration was performed using a state-of-art registration approach. 1H MRI segmentation was performed using primal-dual analysis methods and modern convex optimization techniques with incorporation of region information from 129Xe MRI. We applied the pipeline across a range of pulmonary abnormalities and this computationally efficient pipeline demonstrated high agreement with reference standard, suggesting its suitability for efficient clinical workflows.

Purpose

Key breakthroughs are required to advance novel imaging technologies with respect to research and development, regulatory and patient safety issues and clinical practicability. Here we focus on clinical translation of pulmonary MRI – structure-function methods needed for clinical use. In order to enable clinical translation of pulmonary functional imaging for large-scale, multi-centre and longitudinal applications, we think it is critical to develop image analysis tools that can provide clinically-acceptable and clinically-relevant measurements of regional anatomical and functional abnormalities while integrated into and enhancing clinical workflows. This is important because pulmonary image acquisition is relatively rapid, but image analysis is complex involving multiple modality registration, segmentation and biomarker quantification. Therefore, the objective of this work was to develop a pulmonary segmentation pipeline for quantitative evaluation of 129Xe MRI ventilation across a range of pulmonary abnormalities.

Methods

Subjects and image acquisition:

In total, 20 subjects including eight asthmatics, seven COPD and five healthy subjects provided written informed consent to a study protocol approved by Health Canada. MRI was performed using a whole body 3.0T Discovery MR750 system (General Electric Health Care, Milwaukee, Wisconsin, USA). 1H MRI was performed using a whole-body radio-frequency coil and a 1H fast spoiled gradient-recalled echo (FGRE) sequence using a partial echo (16s acquisition time; repetition time (TR)/echo time (TE)/flip angle=4.3ms/1.2ms/20°; field-of-view (FOV)=40cm×40cm; matrix=128×80; number of slices=14–17; slice thickness=15mm). Hyperpolarized 129Xe MRI static ventilation imaging was performed using a 3D FGRE sequence (acquisition time=16s; TR/TE/flip-angle=7.0ms/1.8ms/variable; FOV=40cm×40cm; matrix=128×128; BW=9kHz; NEX=1; number of slices=16; slice thickness=15mm) during a breath-hold.

1H-129Xe image segmentation:

1H and 129Xe images were resampled to ~3 mm isotropic space using Convert3D1 tools prior to algorithm segmentation. One observer placed seeds on the left and right lung and background on the resampled 1H and 129Xe images three times on 3 different days. A rigid, deformable registration approach2 was used to register the resampled 129Xe and 1H images and the derived deformation field was used to deform the 129Xe seeds to the 1H image space. These seeds were used to estimate the appearance models for the respective regions and jointly formulate the data terms of the Potts approach3 while the regularization terms were generated based on 1H MRI edge information only. The resultant non-convex optimization problem involving binary labeling functions for the three regions was approximated through convex relaxation and further represented by an equivalent max-flow model through variational analysis. 129Xe images were then hierarchically segmented to generate ventilation defects percent (VDP).

Validation:

The accuracy of 1H-129Xe segmentation was evaluated by comparing algorithm lung masks with manual results performed by an experienced observer. We employed the Dice similarity coefficient (DSC), root-mean-squared-error (RMSE) of the Euclidian distance between the two segmented lung surfaces and absolute volume error (|dVe|) to measure the agreement of the two sets of lung volumes. We also compared VDP generated using this new method with a semi-automated approach, previously described.4 Moreover, reproducibility was measured by calculating coefficient of variability (CoV) and intra-class correlation coefficient (ICC) of DSC, RMSE and VDP. The computational efficiency was obtained by averaging the runtime of the repeated algorithm segmentation.

Results

Representative 1H-129Xe segmentation results are shown in Figure 1. The proposed approach yielded a DSC of 89 ±2%, RMSE of 4±1mm and |dVe| of 0.50±0.40L for whole lung. The CoV(ICC) were 0.32%(0.99), 2.4%(0.98), 4%(1.00) for DSC, RMSE and VDP, respectively. Pearson correlation results showed that VDP for five healthy and seven COPD subjects was significantly and strongly correlated with the semi-automated measurements (Figure. 2) with no statistically significant difference (p=0.14). The mean runtime for the proposed approach was ~1 min for each subject, including ~15s for seeding, ~20s for registration, ~5s for max-flow segmentation and ~15s for ventilation defect/VDP generation.

Discussion and Conclusions

The proposed 1H-129Xe segmentation pipeline generates lung volumes and VDP that are in agreement with manual and semi-automated reference standard. This approach involves diminished user interaction and rapid implementation compared to a runtime of ~15min for the semi-automated method. Importantly, the proposed pipeline is highly reproducible in generating lung volumes and providing VDP (CoV of ~4% vs ~20% by the semi-automated approach) from pulmonary 1H and 129Xe MRI. These results suggest that the proposed pipeline might be suitable for large-scale, multi-centre and longitudinal clinical applications.

Acknowledgements

No acknowledgement found.

References

1. Yushkevich et al. Neuroimage 2006.

2. Guo et al. in SPIE Medical Imaging 2015.

3. Yuan et al. in Computer Vision-ECCV 2010.

4. Kirby et al. Academic Radiology 2011.

Figures

Figure 1. 1H-129Xe MRI representative segmentation results. (A) - (C): 1H MRI lung cavity contoured in yellow (left lung) and blue (right lung) from anterior to posterior slices. (D) - (F): 129Xe (hot) registered and overlaid on 1H (gray) with segmented defects (green) in the lung (contours) in anterior-to-posterior slices.

Figure 2. Relationship between the VDP generated by the proposed pipeline (VDPA) and the semi-automated approach (VDPS). Linear correlation (A) and Bland-Altman plots (B) of VDPA and VDPS. Solid and dotted lines represent the mean and the 95% limits of agreement, respectively.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1609