Deformable registration using edge‐preserving scale space for adaptive image‐guided radiation therapy

Li, Dengwang; Wang, Hongjun; Yin, Yong; Wang, Xiuying

doi:10.1120/jacmp.v12i4.3527

Cited by 16 publications

(7 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the past, textureless objects have often been handled by template-matching [11,12]. Image edges then become the dominant information source [21,26], and researchers have developed strategies based on different distances, such as the Hausdorff [15] and the Chamfer [25,13] ones, to match the 3D model against the input image. While effective for poorly-textured objects, these techniques often fail in the presence of mild occlusions and cluttered background.…”

Section: Related Workmentioning

confidence: 99%

Segmentation-Driven 6D Object Pose Estimation

Hugonot

Fua

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

270

242

View full text Add to dashboard Cite

The most recent trend in estimating the 6D pose of rigid objects has been to train deep networks to either directly regress the pose from the image or to predict the 2D locations of 3D keypoints, from which the pose can be obtained using a PnP algorithm. In both cases, the object is treated as a global entity, and a single pose estimate is computed. As a consequence, the resulting techniques can be vulnerable to large occlusions.In this paper, we introduce a segmentation-driven 6D pose estimation framework where each visible part of the objects contributes a local pose prediction in the form of 2D keypoint locations. We then use a predicted measure of confidence to combine these pose candidates into a robust set of 3D-to-2D correspondences, from which a reliable pose estimate can be obtained. We outperform the state-ofthe-art on the challenging Occluded-LINEMOD and YCB-Video datasets, which is evidence that our approach deals well with multiple poorly-textured objects occluding each other. Furthermore, it relies on a simple enough architecture to achieve real-time performance. Input CNNResult

show abstract

Section: Related Workmentioning

confidence: 99%

Segmentation-Driven 6D Object Pose Estimation

Hugonot

Fua

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

270

242

View full text Add to dashboard Cite

show abstract

“…Because of the presence of inconsistent intensities between CT and CBCT, especially in chest and abdomen cases, intensity‐based deformable registration algorithms are susceptible to distorting the tissues, which would cause significant registration error 23, 24. Considering that large gradient generally exists on the border of organs or tissues, gradient field is not subject to the disparity of CBCT and pCT, gradient‐based deformable registration can effectively reduce the aforementioned registration error.…”

Section: Methodsmentioning

confidence: 99%

Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone‐beam CT

Huang

Chen

et al. 2016

J Applied Clin Med Phys

View full text Add to dashboard Cite

Many patients with technically unresectable or medically inoperable hepatocellular carcinoma (HCC) had hepatic anatomy variations as a result of interfraction deformation during fractionated radiotherapy. We conducted this retrospective study to investigate interfractional normal liver dosimetric consequences via reconstructing weekly dose in HCC patients. Twenty‐three patients with HCC received conventional fractionated three‐dimensional conformal radiation therapy (3DCRT) were enrolled in this retrospective investigation. Among them, seven patients had been diagnosed of radiation‐induced liver disease (RILD) and the other 16 patients had good prognosis after treatment course. The cone‐beam CT (CBCT) scans were acquired once weekly for each patient throughout the treatment, deformable image registration (DIR) of planning CT (pCT) and CBCT was performed to acquire modified CBCT (mCBCT), and the structural contours were propagated by the DIR. The same plan was applied to mCBCT to perform dose calculation. Weekly dose distribution was displayed on the pCT dose space and compared using dose difference, target coverage, and dose volume histograms. Statistical analysis was performed to identify the significant dosimetric variations. Among the 23 patients, the three weekly normal liver D50 increased by 0.2 Gy, 4.2 Gy, and 4.7 Gy, respectively, for patients with RILD, and 1.0 Gy, 2.7 Gy, and 3.1 Gy, respectively, for patients without RILD. Mean dose to the normal liver (Dmean) increased by 0.5 Gy, 2.6 Gy, and 4.0 Gy, respectively, for patients with RILD, and 0.4 Gy, 3.1 Gy, and 3.4 Gy, respectively, for patients without RILD. Regarding patients with RILD, the average values of the third weekly D50 and Dmean were both over hepatic radiation tolerance, while the values of patients without RILD were below. The dosimetric consequence showed that the liver dose between patients with and without RILD were different relative to the planned dose, and the RILD patients suffered from liver dose over hepatic radiation tolerance. Evaluation of routinely acquired CBCT images during radiation therapy provides biological information on the organs at risk, and dose estimation based on mCBCT could potentially form the basis for personalized response adaptive therapy.

show abstract

“…Although performing well on textured objects, these methods usually struggle with poorly-textured objects. To deal with this type of objects, template-matching methods try to match the observed object to a stored template [14,15,23,26]. However, they perform badly in the case of partial occlusion or truncation.…”

Section: Related Workmentioning

confidence: 99%

Pose from Shape: Deep Pose Estimation for Arbitrary 3D Objects

Xiao¹,

Qiu²,

Langlois³

et al. 2019

Preprint

View full text Add to dashboard Cite

Most deep pose estimation methods need to be trained for specific object instances or categories. In this work we propose a completely generic deep pose estimation approach, which does not require the network to have been trained on relevant categories, nor objects in a category to have a canonical pose. We believe this is a crucial step to design robotic systems that can interact with new objects "in the wild" not belonging to a predefined category. Our main insight is to dynamically condition pose estimation with a representation of the 3D shape of the target object. More precisely, we train a Convolutional Neural Network that takes as input both a test image and a 3D model, and outputs the relative 3D pose of the object in the input image with respect to the 3D model. We demonstrate that our method boosts performances for supervised category pose estimation on standard benchmarks, namely Pascal3D+, ObjectNet3D and Pix3D, on which we provide results superior to the state of the art. More importantly, we show that our network trained on everyday man-made objects from ShapeNet generalizes without any additional training to completely new types of 3D objects by providing results on the LINEMOD dataset as well as on natural entities such as animals from ImageNet. Our code and model is avalaible at http://imagine.enpc.fr/~xiaoy/PoseFromShape/.

show abstract

Deformable registration using edge‐preserving scale space for adaptive image‐guided radiation therapy

Cited by 16 publications

References 30 publications

Segmentation-Driven 6D Object Pose Estimation

Segmentation-Driven 6D Object Pose Estimation

Investigation of dosimetric variations of liver radiotherapy using deformable registration of planning CT and cone‐beam CT

Pose from Shape: Deep Pose Estimation for Arbitrary 3D Objects

Contact Info

Product

Resources

About