Author: Katrinus Keijnemans, Eric S. Paulson, An Tai 👨🔬
Affiliation: Department of Radiotherapy, University Medical Center Utrecht, Department of Radiation Oncology, Medical College of Wisconsin 🌍
Purpose: Time-weighted mid-position images (MidPs), derived from four-dimensional computed tomography (4DCT) via deformable image registration (DIR), are frequently selected for treatment planning due to their superior image quality and congruence with moving targets on cone-beam CT images used for image-guided radiation therapy (IGRT). This study aims to enhance the quality of MidPs for thoracic tumors using multi-modality DIR (MMDIR) to improve the accuracy of tumor densities (Hounsfield units) on MidPs.
Methods: A MATLAB extension was developed to replace intensity-based DIR (IBDIR) with MMDIR in a MIM (MIM software Inc., OH) workflow used to generate MidPs. 4DCT datasets of 10 lung tumors, acquired using a Siemens Drive scanner and phase-sorted into 10 phases, were used to create MidPs with both algorithms. First, DIR mapped the correspondence between the end-exhale phase and all other phases, generating 3D deformation vectors fields (DVFs) per voxel. Second, the mean DVF was subtracted from the individual DVFs. Third, these DVFs transformed all phases to the mid-position anatomy, and the MidP was derived by averaging over the warped voxel intensities. Mean Hounsfield units within tumor contours on MidPs were compared with those from single-phase images that were closest to the MidPs based on tumor positions.
Results: Tumor density on MidPs generated with IBDIR and MMDIR were comparable and accurate for tumors with superior-inferior motion below 2 cm. However, for tumors with superior-inferior motion above 2 cm, IBDIR-generated MidPs showed substantially lower tumor densities. For the patient with the largest density discrepancy in this study, dose coverage for 95% of the tumor volume was reduced by 3.5% when a plan was calculated on the IBDIR-generated MidPs.
Conclusion: MMDIR should be used for MidP generation in cases of large tumor motion to ensure accurate tumor density, critical for precise dose calculation and improved tumor visibility in IGRT.