Biomechanical deformable image registration of longitudinal lung CT images using vessel information

Cazoulat, Guillaume; Owen, Dawn; Matuszak, Martha M.; Balter, James M.; Brock, Kristy K.

doi:10.1088/0031-9155/61/13/4826

Cited by 32 publications

(45 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, the “Lung + ROI” option uses the same correlation coefficient as for the “Lung” setting, plus controlling contours to penalize contour variations between the registered images. One participant also used a custom version of the MORFEUS algorithm that incorporates boundary conditions on the lung vessel tree …”

Section: Resultsmentioning

confidence: 99%

“…One participant also used a custom version of the MORFEUS algorithm that incorporates boundary conditions on the lung vessel tree. 38 Where the DIR cost function incorporated image similarity metrics, these were based on the intensity mean square error (MSE), cross correlation (CC), squared sum of tissue volume differences (SSTVD), or mutual information (MI). All of the DIR methods used some form of motion field regularization to avoid nonphysical folding of tissue (i.e., negative values of the Jacobian determinant), and a majority of DIR methods also used a "lung focus" (that is, where the DIR optimizer focuses on the lung voxels and/or lung contours).…”

Section: A Summary Of the Ctvi And Dir Motion Field Submissionsmentioning

confidence: 99%

See 1 more Smart Citation

The VAMPIRE challenge: A multi‐institutional validation study of CT ventilation imaging

et al. 2019

Self Cite

View full text Add to dashboard Cite

Purpose CT ventilation imaging (CTVI) is being used to achieve functional avoidance lung cancer radiation therapy in three clinical trials (NCT02528942, NCT02308709, NCT02843568). To address the need for common CTVI validation tools, we have built the Ventilation And Medical Pulmonary Image Registration Evaluation (VAMPIRE) Dataset, and present the results of the first VAMPIRE Challenge to compare relative ventilation distributions between different CTVI algorithms and other established ventilation imaging modalities. Methods The VAMPIRE Dataset includes 50 pairs of 4DCT scans and corresponding clinical or experimental ventilation scans, referred to as reference ventilation images (RefVIs). The dataset includes 25 humans imaged with Galligas 4DPET/CT, 21 humans imaged with DTPA‐SPECT, and 4 sheep imaged with Xenon‐CT. For the VAMPIRE Challenge, 16 subjects were allocated to a training group (with RefVI provided) and 34 subjects were allocated to a validation group (with RefVI blinded). Seven research groups downloaded the Challenge dataset and uploaded CTVIs based on deformable image registration (DIR) between the 4DCT inhale/exhale phases. Participants used DIR methods broadly classified into B‐splines, Free‐form, Diffeomorphisms, or Biomechanical modeling, with CT ventilation metrics based on the DIR evaluation of volume change, Hounsfield Unit change, or various hybrid approaches. All CTVIs were evaluated against the corresponding RefVI using the voxel‐wise Spearman coefficient rS, and Dice similarity coefficients evaluated for low function lung (DSClow) and high function lung (DSChigh). Results A total of 37 unique combinations of DIR method and CT ventilation metric were either submitted by participants directly or derived from participant‐submitted DIR motion fields using the in‐house software, VESPIR. The rS and DSC results reveal a high degree of inter‐algorithm and intersubject variability among the validation subjects, with algorithm rankings changing by up to ten positions depending on the choice of evaluation metric. The algorithm with the highest overall cross‐modality correlations used a biomechanical model‐based DIR with a hybrid ventilation metric, achieving a median (range) of 0.49 (0.27–0.73) for rS, 0.52 (0.36–0.67) for DSClow, and 0.45 (0.28–0.62) for DSChigh. All other algorithms exhibited at least one negative rS value, and/or one DSC value less than 0.5. Conclusions The VAMPIRE Challenge results demonstrate that the cross‐modality correlation between CTVIs and the RefVIs varies not only with the choice of CTVI algorithm but also with the choice of RefVI modality, imaging subject, and the evaluation metric used to compare relative ventilation distributions. This variability may arise from the fact that each of the different CTVI algorithms and RefVI modalities provides a distinct physiologic measurement. Ultimately this variability, coupled with the lack of a “gold standard,” highlights the ongoing importance of further validation studies before CTVI can be widely translated from academic ce...

show abstract

Section: Resultsmentioning

confidence: 99%

Section: A Summary Of the Ctvi And Dir Motion Field Submissionsmentioning

confidence: 99%

The VAMPIRE challenge: A multi‐institutional validation study of CT ventilation imaging

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…This error measurement could be improved when applying this registration method to scans with improved slice thickness, in part because the reproducibility of manually selecting bifurcation points is dependent on slice thickness. Additionally, the modeling of the complex deformations observed in the liver might be improved by using corresponding vessel positions on the two CT scans as additional boundary conditions in the biomechanical model, as previously demonstrated in the lung [35]. However, this would rely on high-quality CT image contrast enabling visualization of vessels and the assumption of vessel preservation, which may not be a good assumption in the presence of large tumor response or cases of tumor induced thrombosis.…”

Section: Discussionmentioning

confidence: 99%

Implementing Radiation Dose-Volume Liver Response in Biomechanical Deformable Image Registration

Polan

Feng

Lawrence

et al. 2017

International Journal of Radiation Oncology*Biology*Physics

View full text Add to dashboard Cite

Purpose Understanding anatomical and functional changes in the liver resulting from radiotherapy is fundamental to the improvement of normal tissue complication models needed to advance personalized medicine. The ability to link pre-treatment and post-treatment imaging is often compromised by significant dose-dependent volumetric changes within the liver that are currently not accounted for in deformable image registration (DIR) techniques. This study investigates using delivered dose, in combination with other patient factors, to biomechanically model longitudinal changes in liver anatomy for follow-up care and retreatment planning. Methods Population models describing the relationship between dose and hepatic volume response were produced using retrospective data from 33 patients treated with focal radiation therapy. A DIR technique was improved by implementing additional boundary conditions associated with the dose-volume response in series with a previously developed biomechanical DIR algorithm. Evaluation of this DIR technique was performed on computed tomography imaging from seven patients by comparing the model-predicted volumetric change within the liver to the observed change, tracking vessel bifurcations within the liver through the deformation process, then determining target registration error (TRE) between the predicted and identified post-treatment bifurcation points. Results Evaluation of the proposed DIR technique showed that all lobes were volumetric deformed to within the respective contour variability of each lobe. The average TRE achieved was 7.3 mm (2.8 mm LR and AP, 5.1 mm SI), with the SI component within the average limiting slice thickness (6.0 mm). This represented a significant improvement (Wilcoxon, p < 0.01) over the application of the previously published biomechanical DIR algorithm (10.9 mm). Conclusion This study demonstrates the feasibility of implementing dose-driven volumetric response in deformable registration, enabling improved accuracy of modeling liver anatomy changes, which could allow for improved dose accumulation, particularly for patients who require additional liver radiotherapy.

show abstract

“…The MDLM method could make robust probabilistic predictions on the occurrence of pulmonary vessels by feeding the learned voxel‐wise features to a logistic regression classifier. The pulmonary vasculature structures could provide abundant information about the vessels and reflect the motion of the lung accurately . We adopted the Harris‐Stephens algorithm to detect feature points on the vasculature probability maps.…”

Section: Methodsmentioning

confidence: 99%

“…On the basis of our experimental experiences, we have observed that landmarks detected using SIFT or Harris corner key point detection methods may have positional uncertainties since these landmarks were detected as the local maximum points in the scale space and were not necessarily stable. On the other hand, the pulmonary vascular structures that have been used for lung image registrations were stable across the respiratory cycle . This was why we have chosen to detect feature points on the vasculature probability maps instead of the original CT images.…”

Section: Methodsmentioning

confidence: 99%

Automatic large quantity landmark pairs detection in 4DCT lung images

Thomas

et al. 2019

Medical Physics

View full text Add to dashboard Cite

Purpose To automatically and precisely detect a large quantity of landmark pairs between two lung computed tomography (CT) images to support evaluation of deformable image registration (DIR). We expect that the generated landmark pairs will significantly augment the current lung CT benchmark datasets in both quantity and positional accuracy. Methods A large number of landmark pairs were detected within the lung between the end‐exhalation (EE) and end‐inhalation (EI) phases of the lung four‐dimensional computed tomography (4DCT) datasets. Thousands of landmarks were detected by applying the Harris‐Stephens corner detection algorithm on the probability maps of the lung vasculature tree. A parametric image registration method (pTVreg) was used to establish initial landmark correspondence by registering the images at EE and EI phases. A multi‐stream pseudo‐siamese (MSPS) network was then developed to further improve the landmark pair positional accuracy by directly predicting three‐dimensional (3D) shifts to optimally align the landmarks in EE to their counterparts in EI. Positional accuracies of the detected landmark pairs were evaluated using both digital phantoms and publicly available landmark pairs. Results Dense sets of landmark pairs were detected for 10 4DCT lung datasets, with an average of 1886 landmark pairs per case. The mean and standard deviation of target registration error (TRE) were 0.47 ± 0.45 mm with 98% of landmark pairs having a TRE smaller than 2 mm for 10 digital phantom cases. Tests using 300 manually labeled landmark pairs in 10 lung 4DCT benchmark datasets (DIRLAB) produced TRE results of 0.73 ± 0.53 mm with 97% of landmark pairs having a TRE smaller than 2 mm. Conclusion A new method was developed to automatically and precisely detect a large quantity of landmark pairs between lung CT image pairs. The detected landmark pairs could be used as benchmark datasets for more accurate and informative quantitative evaluation of DIR algorithms.

show abstract

Biomechanical deformable image registration of longitudinal lung CT images using vessel information

Cited by 32 publications

References 39 publications

The VAMPIRE challenge: A multi‐institutional validation study of CT ventilation imaging

The VAMPIRE challenge: A multi‐institutional validation study of CT ventilation imaging

Implementing Radiation Dose-Volume Liver Response in Biomechanical Deformable Image Registration

Automatic large quantity landmark pairs detection in 4DCT lung images

Contact Info

Product

Resources

About