Generalized Linear Binning to Compare Hyperpolarized 129Xe Ventilation Maps Derived from 3D Radial Gas Exchange Versus Dedicated Multislice Gradient Echo MRI

Rationale: Hyperpolarized Xe-129 ventilation MRI is typically acquired using multislice fast gradient recalled echo (GRE), but interleaved 3D radial Xe-129 gas transfer MRI now provides dissolved-phase and ventilation images from a single breath. To investigate whether these ventilation images provide equivalent quantitative metrics, we introduce generalized linear binning analysis. Methods: This study included 36 patients who had undergone both multislice GRE ventilation and 3D radial gas exchange imaging. Images were then quantified by linear binning to classify voxels into one of four clusters: ventilation defect percentage (VDP), Low-, Medium- or High-ventilation percentage (LVP, MVP, HVP). For 3D radial images, linear binning thresholds were generalized using a Box-Cox rescaled reference histogram. We compared the cluster populations from the two ventilation acquisitions both numerically and spatially. Results: Interacquisition Bland-Altman limits of agreement for the clusters between 3D radial vs GRE were (similar to 7% to 5%) for VDP, (similar to 10% to 14%) for LVP, and (similar to 8% to 8%) for HVP. While binning maps were qualitatively similar between acquisitions, their spatial overlap was modest for VDP (Dice = 0.5 +/- 0.2), and relatively poor for LVP (0.3 +/- 0.1) and HVP (0.2 +/- 0.1). Conclusion: Both acquisitions yield reasonably concordant VDP and qualitatively similar maps. However, poor regional agreement (Dice) suggests that the two acquisitions cannot yet be used interchangeably. However, further improvements in 3D radial resolution and reconciliation of bias field correction may well obviate the need for a dedicated ventilation scan in many cases.