Volume 36, Issue 6, June 2009
Index of content:
- Joint Imaging/Therapy Scientific Session: Room 213A
- Target and Normal Organ Definition
TH‐D‐213A‐01: An Evaluation of FDG‐PET Uptake Thresholds for Head & Neck Target Definition Based On Local Regions of High Inter‐Observer Concordance36(2009); http://dx.doi.org/10.1118/1.3182712View Description Hide Description
Purpose: Thresholding of fluorodeoxyglucose (FDG) uptake imaged with positron emission tomography(PET) has been proposed for radiotherapy target delineation. However, there exists no consensus on the threshold best corresponding to disease extent, due to a lack of truth for validation. The purpose of this investigation was to establish criteria under which a fixed FDG‐PET threshold may be useful for target delineation. An image‐based surrogate of target definition ‘truth’ is proposed, derived from a measure of the concordance between multiple expert observers defining the gross tumor volume (GTV) with FDG‐PET and CT. A method was developed that can automatically localize these regions and was used to evaluate the coincident FDG uptake threshold. Methods and Materials: 10 patients with head and neck (H&N) cancers underwent FDG‐PET‐CT imaging using a hybrid scanner (Biograph 16, Siemens Medical Systems). For each patient, 8 observers specializing in H&N cancers delineated the GTV on concurrently‐displayed FDG‐PET‐CT. Regions of high target definition concordance were localized by forming the union of observer target definitions, applying a gradient filter and selecting an intensity threshold equivalent to 6 of 8 observers. At regions of high concordance not attributable exclusively to CT information, the FDG uptake threshold relative to maximum uptake in the GTV was measured. Results: In 4 patients, the mean FDG uptake threshold was 40.9%. In regions of concordance not attributable to CT, the mean threshold in 6 patients was 32.8%. Conclusion: A method was developed for localizing regions of high inter‐observer concordance. These regions served as a surrogate truth for disease extent against which FDG‐PET uptake thresholds were evaluated. For primary H&N cancers, a threshold of approximately 30–40% was able to delineate the tumor at regions of the target boundary where CT information was not definitive. The application of uniform thresholds for H&N target delineation is not recommended.
TH‐D‐213A‐02: An Improved Iterative Thresholding Approach for 3D PET Tumor Delineation: Phantom Study36(2009); http://dx.doi.org/10.1118/1.3182714View Description Hide Description
Purpose: To develop an improved iterative thresholding method (IITM) for 3D PET tumor delineation, motivated by the iterative thresholding method (ITM). Method and Materials: The IITM method is an iterative thresholding method that selects the threshold level to segment a tumor based on calibrated source‐to‐background threshold‐volume (SBTV) curves. From an initial tumor volume V 0, the source‐to‐background (S/B) ratio R 0 is calculated as the ratio of the average intensity of V 0 and the average intensity of regions of interest in the background. A calibrated SBTV curve corresponding to R 0 is generated by interpolation using those pre‐defined calibrated SBTV curves. This SBTV curve is used for finding a threshold level T 1. The threshold level T 1 is applied to the image and a new estimation of volume V 1 is obtained. This iterative process of updating the SBTV curve, the threshold level and the volume is repeated until the threshold values T i and T i‐1 are not significantly different. In comparison, the ITM method calculates the source as the average intensity in a small neighborhood of the voxel with the maximum intensity and the S/B ratio is fixed during its iterative process. A phantom study using six spherical tumor phantoms (0.5 – 20 mL) with S/B ratio ranging from 16 to 0.5 is used to test these two methods. Also a 3D Gaussian function was used to generate heterogeneous pattern in the phantoms. Results: The IITM method yielded better or equivalent delineation accuracy than the ITM method on all images, particularly at low S/B ratio of 3 for heterogeneous targets. Conclusion: By measuring the source intensity in the entire tumor region and iteratively updating the SBTV curve in addition to the volume and threshold level, the IITM method shows some advantages over the ITM method, particularly for heterogeneous targets.
TH‐D‐213A‐03: Physiological Validation of 4D‐CT‐Based Ventilation Imaging in Patients with Chronic Obstructive Pulmonary Disease (COPD)36(2009); http://dx.doi.org/10.1118/1.3182715View Description Hide Description
Purpose: Four‐dimensional (4D) CT‐based pulmonary ventilation imaging is a new technique and has advantages in speed, resolution, cost and availability. However, the physiological accuracy has not been validated, especially for regional ventilation. The purpose of this study was to validate 4D‐CT‐based ventilation imaging by comparison with distribution and progression of emphysema. Method and Materials: The 4D‐CT images were acquired for radiotherapytreatment planning purposes using the GE multislice PET/CT scanner with the Varian RPM system. Ventilation was evaluated for five patients with chronic obstructive pulmonary disease (COPD) and lungcancer by calculating peak‐inhale to peak‐exhale displacement vector fields (DVFs), and then ventilation based on the Jacobian determinant of deformation. To derive DVFs, we used two fundamentally different deformable image registration (DIR) algorithms: surface‐based registration (DIR s ) and volume‐based registration (DIR v ), of which the geometric accuracy has been validated. The peak‐inhale 4D‐CT images were used for emphysema quantification. Emphysematous lungs were detected by the density masking technique. The 15th percentile point was used as a measure of emphysema progression. Results: There were large discrepancies between distributions of emphysema and ventilation, and between ventilation calculated by two different algorithms. The mean relative volumes of emphysema outside poor‐ventilated lungs were 0.665 (range, 0.520–0.901) and 0.709 (range, 0.464–0.933) for DIR v and DIR v , respectively. There were very weak (−0.013) to strong (−0.763) negative correlations between emphysema progression and the mean ventilation. The use of an algorithm parameter with more or less elasticity for DIR v led to improved coverage of emphysema as well as correlations. Conclusion: This work represents the first physiological validation of 4D‐CT‐based ventilation imaging. Our results demonstrated large discrepancies between ventilation and emphysema. However, a trend toward negative correlation and effects of algorithm parameters indicated promise for this technique. Conflict of Interest: SK, JB, TK, TB and CL are employees of Philips Research Europe.
TH‐D‐213A‐04: Application of Supervised Spectral Clustering for PET Tumor Delineation: A Phantom Study36(2009); http://dx.doi.org/10.1118/1.3182716View Description Hide Description
Purpose:Spectralclustering is a powerful technique for image segmentation. This study investigated the feasibility of using the supervised spectralclustering technique for PETtumor delineation. Method and Materials: A phantom was constructed with six spheres filled with featuring simulated tumors of varying size ranged from 0.5ml to 20ml situated in a cylindrical container filled with ‐FDG. The phantom was scanned for 120 minutes in a PET/CT scanner. For every 2 minute scan, 6 images were reconstructed by using the iterative OSEM algorithm with 5mm‐FWHM Gaussian smoothing filter and followed with an anisotropic filtering. The supervised spectralclustering algorithm using normalized cut was then applied to segment the filtered images.Spectralclustering uses the eigenvectors and eigenvalues of a similarity matrix (based on both proximity and intensity) to partition pixels into clusters. Its segmentation performance was evaluated with two metrics: area detection error and area overlap metric. Results: The supervised spectralclustering technique partitions the image to a designated number of regions according the measure of the pair‐wise affinities between pixels. The spectralclustering has a rather steady segmentation performance of greater than 85% area overlap for images with S/B ratios ⩾ 5. The detected area is enclosed in the true area until the S/B ratio drops to the lowest level of 2. Conclusion: The preliminary results demonstrate the potential of the supervised spectralclustering method for PETtumor segmentation. Compared with other PETtumor delineation techniques, it segments the PETimage with minimal human intervention and is instrument independent.
TH‐D‐213A‐05: A Method to Estimate 4DCT Deformable Registration Errors Via Principal Components Analysis36(2009); http://dx.doi.org/10.1118/1.3182717View Description Hide Description
Purpose: To extract continuous deformable image registration error maps from 4DCTs using a small set of ground truth landmarks. Method and Materials: Assuming that we have a set of displacement vector fields (DVFs) calculated via deformable registration from 4DCT breathing phases, and a small set of validation landmarks, we use principal components analysis (PCA) to separate the eigenmodes of breathing from the error eigenmodes. We then reconstruct the DVF error maps from their principal components. To test this method we made a 2D model of breathing motion in the thorax with a 16×16 pixel DVF to which we added simulated errors. We selected 20 pixels as ground truth landmarks and applied our PCA method to the simulated data to estimate the error maps. We applied the same algorithm to the Point‐validated Pixel‐base (POPI) 4DCT data set, which consists of a 4D CT in which clinicians have located 41 point landmarks in 10 reconstructed CT breathing phases. Results: In the numerical simulations our method successfully recovered the artificial DVF error map although the error amplitudes were underestimated by 10–30%. The 4DCT analysis showed PCA eigenvalue spectra that are consistent with the assumptions of our error estimation method and also indicated that the landmarks provided by POPI model are sufficiently representative to be used as ground truth for our analysis. Conclusion: Point‐by‐point landmark validation of deformable image registration results can give highly selective results that do not realistically reflect the true spatial distribution of registration errors. Our PCA based method allows a more realistic picture of image registration and 4DCT motion mapping errors.
This research was supported in part by NIH grant P01CA116602.
36(2009); http://dx.doi.org/10.1118/1.3182718View Description Hide Description
Purpose: To develop a high performance registration method for mapping objects of interest in treatment images to the 3D image domain of the planning CT in IGRT for improvement of treatment accuracy and potentially adaptive online radiotherapy.Method and Materials: A meshless deformable model has been adopted in a registration framework. The target and reference images are first registered rigidly using bony structures. Then the reference image is sampled around the volume of interest and the adjacent critical functioning organs to form a meshless point cloud, which deforms under the influence of internal structural constraints as well as intensity differences between target and reference images. The meshless model acted recursively with traditional image feature based registration to determine the optimal mapping between the reference and the target. The method was evaluated on 15 prostate datasets (each dataset includes CT and CBCT of one patient). The structures of interest in all images were delineated by a radiation oncologist (serving as the gold standard) in our new validation framework. We evaluated our method by quantitatively measuring the convergence of critical clinical objects and major image features in the reference and registered target image.Results: For all 15 datasets, the new non‐rigid registration algorithm can build the displacement map between CT and CBCTimages cropped around the prostate within an average time of 38.2(±2.9) seconds. The average volumetric similarity between the registered and the reference object are over 92%. Average distance between the registered and the reference surface is approximately 1.3mm with a maximum error under 4 mm. Conclusion: The meshless method reduces the degree of freedom in the solution space and significantly enhances the registration speed without loss of accuracy. Furthermore, the established meshless framework minimizes manual intervention during the registration thereby enhancing the potential for efficient clinical implementations.
TH‐D‐213A‐07: A Novel Inverse‐Consistent Feature‐Based Non‐Rigid Registration Method That Improves the Mapping of Organs with Large‐Scale Deformations36(2009); http://dx.doi.org/10.1118/1.3182719View Description Hide Description
Purpose: Unidirectional feature‐based registration methods can result in inconsistent correspondence between the forward and the backward transformation and in incoherent anatomical mapping. The aim of this work is to develop, test and validate an inverse‐consistent feature‐based non‐rigid registration method in order to improve the registration of organs that exhibit large deformations. Methods and Materials: Thin Plate Splines Robust Point Matching (TPS‐RPM) is a unidirectional algorithm that iteratively calculates the correspondence and the transformation between two point sets (e.g., anatomical structures,organs). An inverse‐consistent version of TPS‐RMP (IC‐TPS‐RPM) was developed that jointly estimates the forward and the backward transformations and that uses both transformations to determine the correspondence. IC‐TPS‐RPM was compared with TPS‐RPM by registering organs with large deformations in five patients. For each patient the contoured cervix‐uteri and bladders on a series of three variable bladder filling CT‐scans (empty to full) were employed. The mean ratio between the volume of full bladder and the volume of empty bladder was 5.7. The registration accuracy error, the inverse‐consistency error, the residual distances after transforming anatomical landmarks and the registration time were calculated using both algorithms. Results: The registrations performed with IC‐TPS‐RPM have on average 10% and 70% better accuracy and inverse‐consistency, respectively when compared with the non‐symmetric TPS‐RPM. By using IC‐TPR‐RPM the residual distances after transforming anatomical landmarks for the registration of full to empty bladder reduced by 47% and by 11% for all landmarks. Moreover, the registration time for computing the forward and the backward transformations decreased by 29%. Conclusions: Compared with TPS‐RPM the new IC‐TPS‐RPM method improves the registration accuracy, the inverse‐consistency and the anatomical correspondence. For cases with large deformations accurate transformations were obtained with IC‐TPS‐RPM, while TPS‐RPM failed. Furthermore, IC‐TPS‐RPM requires less time to compute the forward and the backward transformations.