Deformable image registration (DIR) is emerging as a tool in radiation therapy for calculating the cumulative dose distribution across multiple fractions of treatment. Unfortunately, due to the variable nature of DIR algorithms and dependence of performance on image quality, registration errors can result in dose accumulation errors. In this study, landmarked images were used to characterize the DIR error throughout an image space and determine its impact on dosimetric analysis. Ten thoracic 4DCT images with 300 landmarks per image study matching the end-inspiration and end-expiration phases were obtained from 'dir-labs'. DIR was performed using commercial software MIM Maestro. The range of dose uncertainty (RDU) was calculated at each landmark pair as the maximum and minimum of the doses within a sphere around the landmark in the end-expiration phase. The radius of the sphere was defined by a measure of DIR error which included either the actual DIR error, mean DIR error per study, constant errors of 2 or 5 mm, inverse consistency error, transitivity error or the distance discordance metric (DDM). The RDUs were evaluated using the magnitude of dose uncertainty (MDU) and inclusion rate (IR) of actual error lying within the predicted RDU. The RDU was calculated for 300 landmark pairs on each 4DCT study for all measures of DIR error. The most representative RDU was determined using the actual DIR error with a MDU of 2.5 Gy and IR of 97%. Across all other measures of DIR error, the DDM was most predictive with a MDU of 2.5 Gy and IR of 86%, closest to the actual DIR error. The proposed method represents the range of dosimetric uncertainty of DIR error using either landmarks at specific voxels or measures of registration accuracy throughout the volume.
Deformable image registration (DIR) has become a common tool in medical imaging across both diagnostic and treatment specialties, but the methods used offer varying levels of accuracy. Evaluation of DIR is commonly performed using manually selected landmarks, which is subjective, tedious and time consuming. We propose a semi-automated method that saves time and provides accuracy comparable to manual selection. Three landmarking methods including manual (with two independent observers), scale invariant feature transform (SIFT), and SIFT with manual editing (SIFT-M) were tested on 10 thoracic 4DCT image studies corresponding to the 0% and 50% phases of respiration. Results of each method were evaluated against a gold standard (GS) landmark set comparing both mean and proximal landmark displacements. The proximal method compares the local deformation magnitude between a test landmark pair and the closest GS pair. Statistical analysis was done using an intra class correlation (ICC) between test and GS displacement values. The creation time per landmark pair was 22, 34, 2.3, and 4.3 s for observers 1 and 2, SIFT, and SIFT-M methods respectively. Across 20 lungs from the 10 CT studies, the ICC values between the GS and observer 1 and 2, SIFT, and SIFT-M methods were 0.85, 0.85, 0.84, and 0.82 for mean lung deformation, and 0.97, 0.98, 0.91, and 0.96 for proximal landmark deformation, respectively. SIFT and SIFT-M methods have an accuracy that is comparable to manual methods when tested against a GS landmark set while saving 90% of the time. The number and distribution of landmarks significantly affected the analysis as manifested by the different results for mean deformation and proximal landmark deformation methods. Automatic landmark methods offer a promising alternative to manual landmarking, if the quantity, quality and distribution of landmarks can be optimized for the intended application.
BackgroundHead and neck cancers are commonly treated with radiation therapy, but due to possible volume changes, plan adaptation may be required during the course of treatment. Currently, plan adaptations consume significant clinical resources. Existing methods to evaluate the need for plan adaptation requires deformable image registration (DIR) to a new CT simulation or daily cone beam CT (CBCT) images and the recalculation of the dose distribution. In this study, we explore a tool to assist the decision for plan adaptation using a CBCT without re‐computation of dose, allowing for rapid online assessment.MethodsThis study involved 18 head and neck cancer patients treated with CBCT image guidance who had their treatment plan modified based on a new CT simulation (ReCT). Dose changes were estimated using different methods and compared to the current gold standard of using DIR between the planning CT scan (PCT) and ReCT with recomputed dose. The first and second methods used DIR between the PCT and daily CBCT with the planned dose or recalculated dose from the ReCT respectively, with the dose transferred to the CBCT using rigid registration. The necessity of plan adaptation was assessed by the change in dose to 95% of the planning target volume (D95) and mean dose to the parotids.ResultsThe treatment plans were adapted clinically for all 18 patients but only 7 actually needed an adaptation yielding 11 unnecessary adaptations. Applying a method using the daily CBCT with the planned dose distribution would have yielded only four unnecessary adaptations and no missed adaptations: a significant improvement from that done clinically.ConclusionUsing the DIR between the planning CT and daily CBCT can flag cases for plan adaptation before every fraction while not requiring a new re‐planning CT scan and dose recalculation.
The developed models can potentially guide dose selection and provide prognostic information but still require external validation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.