Dataset Information

A neural network-based 2D/3D image registration quality evaluator for pediatric patient setup in external beam radiotherapy.

ABSTRACT: Our purpose was to develop a neural network-based registration quality evaluator (RQE) that can improve the 2D/3D image registration robustness for pediatric patient setup in external beam radiotherapy. Orthogonal daily setup X-ray images of six pediatric patients with brain tumors receiving proton therapy treatments were retrospectively registered with their treatment planning computed tomography (CT) images. A neural network-based pattern classifier was used to determine whether a registration solution was successful based on geometric features of the similarity measure values near the point-of-solution. Supervised training and test datasets were generated by rigidly registering a pair of orthogonal daily setup X-ray images to the treatment planning CT. The best solution for each registration task was selected from 50 optimizing attempts that differed only by the randomly generated initial transformation parameters. The distance from each individual solution to the best solution in the normalized parametrical space was compared to a user-defined error tolerance to determine whether that solution was acceptable. A supervised training was then used to train the RQE. Performance of the RQE was evaluated using test dataset consisting of registration results that were not used in training. The RQE was integrated with our in-house 2D/3D registration system and its performance was evaluated using the same patient dataset. With an optimized sampling step size (i.e., 5 mm) in the feature space, the RQE has the sensitivity and the specificity in the ranges of 0.865-0.964 and 0.797-0.990, respectively, when used to detect registration error with mean voxel displacement (MVD) greater than 1 mm. The trial-to-acceptance ratio of the integrated 2D/3D registration system, for all patients, is equal to 1.48. The final acceptance ratio is 92.4%. The proposed RQE can potentially be used in a 2D/3D rigid image registration system to improve the overall robustness by rejecting unsuccessful registration solutions. The RQE is not patient-specific, so a single RQE can be constructed and used for a particular application (e.g., the registration for images acquired on the same anatomical site). Implementation of the RQE in a 2D/3D registration system is clinically feasible.

SUBMITTER: Wu J

PROVIDER: S-EPMC5690212 | biostudies-literature | 2016 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A neural network-based 2D/3D image registration quality evaluator for pediatric patient setup in external beam radiotherapy.

Wu Jian J Su Zhong Z Li Zuofeng Z

Journal of applied clinical medical physics 20160108 1

Our purpose was to develop a neural network-based registration quality evaluator (RQE) that can improve the 2D/3D image registration robustness for pediatric patient setup in external beam radiotherapy. Orthogonal daily setup X-ray images of six pediatric patients with brain tumors receiving proton therapy treatments were retrospectively registered with their treatment planning computed tomography (CT) images. A neural network-based pattern classifier was used to determine whether a registration ...[more]

PMID: 26894329

Similar Datasets

Project description:BackgroundSingle-isocenter (SI) noncoplanar volumetric modulated arc therapy (NC-VMAT) has been widely used in stereotactic radiosurgery (SRS) or hypofractionated stereotactic radiotherapy (HSRT) for multiple brain metastases (BMs). However, it is critical to verify patient positioning at a noncoplanar couch angle. This study aimed to compare the noncoplanar setup discrepancies between kilo-voltage/mega-voltage image (kV/MV) orthogonal image pairs with a 2-dimensional/3-dimensional (2D/3D) matching mode and noncoplanar cone-beam computed tomography (NC-CBCT) with a 3D/3D matching mode in SI NC-VMAT HSRT for multiple BMs.MethodsTwenty patients with multiple BMs [2-5] who underwent SI NC-VMAT HSRT were enrolled in this study. Prior to each noncoplanar field delivery, both kV/MV orthogonal image pairs and NC-CBCT were used to determine setup errors. The setup error values reported by NC-CBCT were defined as the gold standard and compared to those reported by kV/MV orthogonal image pairs. The Bland-Altman analysis method was utilized to assess the agreement of the two positioning modalities.ResultsIn total, 104 kV/MV image pairs and NC-CBCT scans were acquired. The mean setup error differences (SEDs; absolute values) between the two positioning systems were 0.17 mm, 0.21 mm, 0.16 mm, 0.22°, 0.18°, and 0.17° in the vertical, longitudinal, lateral, yaw, pitch, and roll directions, respectively. The maximum SEDs regarding translation and rotation occurred in the longitudinal and yaw directions at 0.60 mm and 0.8°, respectively. Bland-Altman analysis showed excellent agreement between the two positioning modalities, and the 95% limits of agreement (LOAs) never exceeded 0.6 mm and 0.6° in the translational and rotational directions, respectively. Only 4.80% of SEDs exceeded the tolerance of 0.5 mm/0.5°.ConclusionsOrthogonal kV/MV image pairs with 2D/3D matching mode could provide comparable accuracy for noncoplanar positioning as NC-CBCT with 3D/3D matching mode.

Project description:To develop a neural network based registration quality evaluator (RQE) that can identify unsuccessful 3D/3D image registrations for the head-and-neck patient setup in radiotherapy.A two-layer feed-forward neural network was used as a RQE to classify 3D/3D rigid registration solutions as successful or unsuccessful based on the features of the similarity surface near the point-of-solution. The supervised training and test data sets were generated by rigidly registering daily cone-beam CTs to the treatment planning fan-beam CTs of six patients with head-and-neck tumors. Two different similarity metrics (mutual information and mean-squared intensity difference) and two different types of image content (entire image versus bony landmarks) were used. The best solution for each registration pair was selected from 50 optimizing attempts that differed only by the initial transformation parameters. The distance from each individual solution to the best solution in the normalized parametrical space was compared to a user-defined error threshold to determine whether that solution was successful or not. The supervised training was then used to train the RQE. The performance of the RQE was evaluated using the test data set that consisted of registration results that were not used in training.The RQE constructed using the mutual information had very good performance when tested using the test data sets, yielding the sensitivity, the specificity, the positive predictive value, and the negative predictive value in the ranges of 0.960-1.000, 0.993-1.000, 0.983-1.000, and 0.909-1.000, respectively. Adding a RQE into a conventional 3D/3D image registration system incurs only about 10%-20% increase of the overall processing time.The authors' patient study has demonstrated very good performance of the proposed RQE when used with the mutual information in identifying unsuccessful 3D/3D registrations for daily patient setup. The classifier had very good generality and required only to be trained once for each implementation. When the RQE is incorporated with an automated 3D/3D image registration system, it can improve the robustness of the system.

Project description:PurposeImage-guided navigation and surgical robotics are the next frontiers of minimally invasive surgery. Assuring safety in high-stakes clinical environments is critical for their deployment. 2D/3D registration is an essential, enabling algorithm for most of these systems, as it provides spatial alignment of preoperative data with intraoperative images. While these algorithms have been studied widely, there is a need for verification methods to enable human stakeholders to assess and either approve or reject registration results to ensure safe operation.MethodsTo address the verification problem from the perspective of human perception, we develop novel visualization paradigms and use a sampling method based on approximate posterior distribution to simulate registration offsets. We then conduct a user study with 22 participants to investigate how different visualization paradigms (Neutral, Attention-Guiding, Correspondence-Suggesting) affect human performance in evaluating the simulated 2D/3D registration results using 12 pelvic fluoroscopy images.ResultsAll three visualization paradigms allow users to perform better than random guessing to differentiate between offsets of varying magnitude. The novel paradigms show better performance than the neutral paradigm when using an absolute threshold to differentiate acceptable and unacceptable registrations (highest accuracy: Correspondence-Suggesting (65.1%), highest F1 score: Attention-Guiding (65.7%)), as well as when using a paradigm-specific threshold for the same discrimination (highest accuracy: Attention-Guiding (70.4%), highest F1 score: Corresponding-Suggesting (65.0%)).ConclusionThis study demonstrates that visualization paradigms do affect the human-based assessment of 2D/3D registration errors. However, further exploration is needed to understand this effect better and develop more effective methods to assure accuracy. This research serves as a crucial step toward enhanced surgical autonomy and safety assurance in technology-assisted image-guided surgery.

Project description:To provide a comprehensive assessment of patient setup accuracy in 6 degrees of freedom (DOFs) using 2-dimensional/3-dimensional (2D/3D) image registration with on-board 2-dimensional kilovoltage (OB-2 DkV) radiographic images, we evaluated cranial, head and neck (HN), and thoracic and abdominal sites under clinical conditions. A fast 2D/3D image registration method using graphics processing unit GPU was modified for registration between OB-2 DkV and 3D simulation computed tomography (simCT) images, with 3D/3D registration as the gold standard for 6 DOF alignment. In 2D/3D registration, body roll rotation was obtained solely by matching orthogonal OB-2 DkV images with a series of digitally reconstructed radiographs (DRRs) from simCT with a small rotational increment along the gantry rotation axis. The window/level adjustments for optimal visualization of the bone in OB-2 DkV and DRRs were performed prior to registration. Ideal patient alignment at the isocenter was calculated and used as an initial registration position. In 3D/3D registration, cone-beam CT (CBCT) was aligned to simCT on bony structures using a bone density filter in 6DOF. Included in this retrospective study were 37 patients treated in 55 fractions with frameless stereotactic radiosurgery or stereotactic body radiotherapy for cranial and paraspinal cancer. A cranial phantom was used to serve as a control. In all cases, CBCT images were acquired for patient setup with subsequent OB-2 DkV verification. It was found that the accuracy of the 2D/3D registration was 0.0 ± 0.5 mm and 0.1° ± 0.4° in phantom. In patient, it is site dependent due to deformation of the anatomy: 0.2 ± 1.6 mm and -0.4° ± 1.2° on average for each dimension for the cranial site, 0.7 ± 1.6 mm and 0.3° ± 1.3° for HN, 0.7 ± 2.0 mm and -0.7° ± 1.1° for the thorax, and 1.1 ± 2.6 mm and -0.5° ± 1.9° for the abdomen. Anatomical deformation and presence of soft tissue in 2D/3D registration affect the consistency with 3D/3D registration in 6 DOF: the discrepancy increases in superior to inferior direction.

Project description:Surgical reduction of pelvic dislocation is a challenging procedure with poor long-term prognosis if reduction does not accurately restore natural morphology. The procedure often requires long fluoroscopic exposure times and trial-and-error to achieve accurate reduction. We report a method to automatically compute the target pose of dislocated bones in preoperative CT and provide 3D guidance of reduction using routine 2D fluoroscopy. A pelvic statistical shape model (SSM) and a statistical pose model (SPM) were formed from an atlas of 40 pelvic CT images. Multi-body bone segmentation was achieved by mapping the SSM to a preoperative CT via an active shape model. The target reduction pose for the dislocated bone is estimated by fitting the poses of undislocated bones to the SPM. Intraoperatively, multiple bones are registered to fluoroscopy images via 3D-2D registration to obtain 3D pose estimates from 2D images. The method was examined in three studies: (1) a simulation study of 40 CT images simulating a range of dislocation patterns; (2) a pelvic phantom study with controlled dislocation of the left innominate bone; (3) a clinical case study investigating feasibility in images acquired during pelvic reduction surgery. Experiments investigated the accuracy of registration as a function of initialization error (capture range), image quality (radiation dose and image noise), and field of view (FOV) size. The simulation study achieved target pose estimation with translational error of median 2.3 mm (1.4 mm interquartile range, IQR) and rotational error of 2.1° (1.3° IQR). 3D-2D registration yielded 0.3 mm (0.2 mm IQR) in-plane and 0.3 mm (0.2 mm IQR) out-of-plane translational error, with in-plane capture range of ±50 mm and out-of-plane capture range of ±120 mm. The phantom study demonstrated 3D-2D target registration error of 2.5 mm (1.5 mm IQR), and the method was robust over a large dose range, down to 5 [Formula: see text]Gy/frame (an order of magnitude lower than the nominal fluoroscopic dose). The clinical feasibility study demonstrated accurate registration with both preoperative and intraoperative radiographs, yielding 3.1 mm (1.0 mm IQR) projection distance error with robust performance for FOV ranging from 340 × 340 mm2 to 170 × 170 mm2 (at the image plane). The method demonstrated accurate estimation of the target reduction pose in simulation, phantom, and a clinical feasibility study for a broad range of dislocation patterns, initialization error, dose levels, and FOV size. The system provides a novel means of guidance and assessment of pelvic reduction from routinely acquired preoperative CT and intraoperative fluoroscopy. The method has the potential to reduce radiation dose by minimizing trial-and-error and to improve outcomes by guiding more accurate reduction of joint dislocations.

Dataset Information

A neural network-based 2D/3D image registration quality evaluator for pediatric patient setup in external beam radiotherapy.

Publications

A neural network-based 2D/3D image registration quality evaluator for pediatric patient setup in external beam radiotherapy.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets