Room: AAPM ePoster Library
Purpose: Integrating functional data such as ventilation into radiotherapy planning for lung cancer has the potential to aid clinicians in reducing adverse effects. The aim of this study is to validate a deformable image registration (DIR) technique for assessing lung ventilation using four-dimensional computed tomography (4DCT).
Methods: The 4DCT scans of a patient cohort (n=31) were processed retrospectively with a program called Advanced Normalized Tools registration (ANTs). To assess lung ventilation using a DIR technique, the lung deformation is evaluated between the respiratory extrema. Deformation of the lung has a localized impact on air flow, which is correlated with ventilation. Volume variation ascribed to each voxel is derived from the Jacobian of the deformation field. Validation of the technique was performed using several metrics: the target registration error (TRE) using anatomical markers as well as the evaluation of Dice coefficient (DC) and Hausdorff distance (HD) between the deformed lung contours and the aimed contours. Single-photon emission computed tomography (SPECT) data, the current standard to estimate lung ventilation, is used as a ground truth for comparison with our proposed method.
Results: The distance between five corresponding anatomical landmarks was calculated for six patients. The mean TRE was (9.5 ± 1.0) mm for those six. Additionally, the similarity between the registered and targeted lung contours showed a mean DC of (0.91 ± 0.03) and a mean HD of (8.0 ± 0.9) mm for all 31 patients.
Conclusion: The overall performance of the registration technique will allow us to determine the robustness of our methods in extracting ventilation data from 4DCT scans. Future work will focus on lung functionality based on both ventilation and perfusion data derived from 4DCT and dual-energy CT with iodine contrast, respectively.