Room: Exhibit Hall | Forum 5
Purpose: In radiation therapy, there is growing interest in identifying and preferentially avoiding higher function lung tissue. Previous research considered 4DCT-computed volume changes during breathing to be reflective of ventilation, and therefore as an adequate surrogate for lung function. Here, we examine whether 4DCT-computed ventilation (â€œcomputed ventilationâ€?) is equivalent to ventilation from Â¹Â²â?¹Xe-MR imaging (â€œMR ventilationâ€?).
Methods: From previously-studied patients with Â¹Â²â?¹Xe-MR ventilation scans, two were identified who had 4DCT scans. MIM was used to deformably register from peak inhalation to peak exhalation, from which computed ventilation was derived via the Jacobian determinant of the deformation vector field. MR data, acquired during a 1L inhale from functional residual capacity, were rigidly aligned to the peak-inhale CT volume, allowing for voxel-to-voxel comparison of MR vs. computed ventilation. Both were smoothed with a 1cmÂ³ median box filter. Lung regions were contoured on CTs, and within these contours, isopercentile volumes were compared between MR and computed ventilation.
Results: Comparison of regions containing the lowest 33%, middle 33%, highest 33%, and highest 10% of MR vs. computed ventilation resulted in low dice coefficients for the two patients: 0.31 and 0.43 for the lowest 33% of voxels, 0.31 and 0.32 for the middle 33% of voxels, 0.44 and 0.51 for the highest 33% of voxels, and 0.10 and 0.18 for the highest 10% of voxels. The Pearson correlation coefficients were also low: Ï? = 0.065 and Ï? = 0.297.
Conclusion: Ventilation maps computed from the 4DCT displacement vector fields do not correlate strongly with ventilation measured with Â¹Â²â?¹Xe-MR imaging. Further, regions of interest for functional avoidance planning show limited overlap, which could result in different functional-optimized plans depending on the modality used to measure ventilation. Further study could expand the patient cohort and quantify errors induced in functional planning.
Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by grants R01HL105643 and P41EB015897 National Institutes of Health (NIH). BD is a founder of and shareholder in Polarean Imaging, outside the submitted work, and has a patent (US 9625550 B2) receiving royalties paid by Polarean Imaging.
Not Applicable / None Entered.