W. Curran

WE-EF-210-08: BEST IN PHYSICS (IMAGING): 3D Prostate Segmentation in Ultrasound Images Using Patch-Based Anatomical Feature

Purpose: Transrectal ultrasound (TRUS) is the standard imaging modality for the image-guided prostate-cancer interventions (e.g., biopsy and brachytherapy) due to its versatility and real-time capability. Accurate segmentation of the prostate plays a key role in biopsy needle placement, treatment planning, and motion monitoring. As ultrasound images have a relatively low signal-to-noise ratio (SNR), automatic segmentation of the prostate is difficult. However, manual segmentation during biopsy or radiation therapy can be time consuming. We are developing an automated method to address this technical challenge. Methods: The proposed segmentation method consists of two major stages: the training stage and the segmentation stage. During the training stage, patch-based anatomical features are extracted from the registered training images with patient-specific information, because these training images have been mapped to the new patient’ images, and the more informative anatomical features are selected to train the kernel support vector machine (KSVM). During the segmentation stage, the selected anatomical features are extracted from newly acquired image as the input of the well-trained KSVM and the output of this trained KSVM is the segmented prostate of this patient. Results: This segmentation technique was validated with a clinical study of 10 patients. The accuracy of our approach was assessed using the manual segmentation. The mean volume Dice Overlap Coefficient was 89.7±2.3%, and the average surface distance was 1.52 ± 0.57 mm between our and manual segmentation, which indicate that the automatic segmentation method works well and could be used for 3D ultrasound-guided prostate intervention. Conclusion: We have developed a new prostate segmentation approach based on the optimal feature learning framework, demonstrated its clinical feasibility, and validated its accuracy with manual segmentation (gold standard). This segmentation technique could be a useful tool for image-guided interventions in prostate-cancer diagnosis and treatment. This research is supported in part by DOD PCRP Award W81XWH-13-1-0269, and National Cancer Institute (NCI) Grant CA114313.

TU-F-BRF-02: MR-US Prostate Registration Using Patient-Specific Tissue Elasticity Property Prior for MR-Targeted, TRUS-Guided HDR Brachytherapy

PURPOSE High-dose-rate (HDR) brachytherapy has become a popular treatment modality for prostate cancer. Conventional transrectal ultrasound (TRUS)-guided prostate HDR brachytherapy could benefit significantly from MR-targeted, TRUS-guided procedure where the tumor locations, acquired from the multiparametric MRI, are incorporated into the treatment planning. In order to enable this integration, we have developed a MR-TRUS registration with a patient-specific biomechanical elasticity prior. METHODS The proposed method used a biomechanical elasticity prior to guide the prostate volumetric B-spline deformation in the MRI and TRUS registration. The patient-specific biomechanical elasticity prior was generated using ultrasound elastography, where two 3D TRUS prostate images were acquired under different probe-induced pressures during the HDR procedure, which takes 2-4 minutes. These two 3D TRUS images were used to calculate the local displacement (elasticity map) of two prostate volumes. The B-spline transformation was calculated by minimizing the Euclidean distance between the normalized attribute vectors of the prostate surface landmarks on the MR and TRUS. This technique was evaluated through two studies: a prostate-phantom study and a pilot study with 5 patients undergoing prostate HDR treatment. The accuracy of our approach was assessed through the locations of several landmarks in the post-registration and TRUS images; our registration results were compared with the surface-based method. RESULTS For the phantom study, the mean landmark displacement of the proposed method was 1.29±0.11 mm. For the 5 patients, the mean landmark displacement of the surface-based method was 3.25±0.51 mm; our method, 1.71±0.25 mm. Therefore, our proposed method of prostate registration outperformed the surfaced-based registration significantly. CONCLUSION We have developed a novel MR-TRUS prostate registration approach based on patient-specific biomechanical elasticity prior. Successful integration of multi-parametric MR and TRUS prostate images provides a prostate-cancer map for treatment planning, enables accurate dose planning and delivery, and potentially enhances prostate HDR treatment outcome.

SU-D-9A-06: 3D Localization of Neurovascular Bundles Through MR-TRUS Registration in Prostate Radiotherapy

PURPOSE Erectile dysfunction (ED) is the most common complication of prostate-cancer radiotherapy (RT) and the major mechanism is radiation-induced neurovascular bundle (NVB) damage. However, the localization of the NVB remains challenging. This studys purpose is to accurately localize 3D NVB by integrating MR and transrectal ultrasound (TRUS) images through MR-TRUS fusion. METHODS T1 and T2-weighted MR prostate images were acquired using a Philips 1.5T MR scanner and a pelvic phase-array coil. The 3D TRUS images were captured with a clinical scanner and a 7.5 MHz biplane probe. The TRUS probe was attached to a stepper; the B-mode images were captured from the prostate base to apex at a 1-mm step and the Doppler images were acquired in a 5-mm step. The registration method modeled the prostate tissue as an elastic material, and jointly estimated the boundary condition (surface deformation) and the volumetric deformations under elastic constraint. This technique was validated with a clinical study of 7 patients undergoing RT treatment for prostate cancer. The accuracy of our approach was assessed through the locations of landmarks, as well as previous ultrasound Doppler images of patients. RESULTS MR-TRUS registration was successfully performed for all patients. The mean displacement of the landmarks between the post-registration MR and TRUS images was 1.37±0.42 mm, which demonstrated the precision of the registration based on the biomechanical model; and the NVB volume Dice Overlap Coefficient was 92.1±3.2%, which demonstrated the accuracy of the NVB localization. CONCLUSION We have developed a novel approach to improve 3D NVB localization through MR-TRUS fusion for prostate RT, demonstrated its clinical feasibility, and validated its accuracy with ultrasound Doppler data. This technique could be a useful tool as we try to spare the NVB in prostate RT, monitor NBV response to RT, and potentially improve post-RT potency outcomes.

TH‐C‐217BCD‐02: Ultrasound Texture Analysis of Radiation‐Induced Parotid‐Gland Injury in Post‐Radiotherapy Head‐And‐Neck Patients: Feasibility Study

Purpose: Xerostomia (dry mouth), resulting from radiation damage to the parotid glands, is one of the most common and distressing side effects of head‐and‐neck cancerradiotherapy. We have developed a family of sonographic texture features to evaluate the morphologic and microstructural integrity of the parotid glands, and investigate the feasibility of quantitative evaluation of radiotherapy‐induced parotid‐gland injury. Methods: In this pilot study, 12 post‐radiotherapy head‐and‐neck cancer patients and 7 healthy volunteers were enrolled. Each participant underwent one ultrasound study, and longitudinal (vertical) ultrasound scans were performed of the bilateral parotids. The averaged follow‐up time for the post‐radiotherapy patients was 17.2 months and the median radiationdose was 32.3 Gy. Eight grey level co‐occurrence matrix (GLCM) features were derived from the B‐mode images. The associated parameters for texture features can be summarized as follows: 1. Inverse differential moment (IDM): local homogeneity; 2. Contrast: difference of gray‐scale through continuous pixels of the image; 3. Angular second moment (ASM): homogenous texture; 4. Entropy: disorder of the image. 5. Variance: heterogeneity; 6. Correlation: linear relationship between the gray‐scale of pixel pairs; 7. Cluster shade and 8. Cluster prominence: the perceptual concepts of uniformity and proximity. Results: Significant differences (p<0.05) were observed in 8 sonographic features, between normal and irradiated parotid glands. IDM value decreased from 7.88±1.14E‐2 (normal) to 6.93±1.54E‐2 (irradiated); Contrast value increased from 3.41±1.11E+2 to 8.45±3.46E+2; ASM value decreased from 6.06±1.72E‐4 to 3.02±0.87E‐4; Entropy value increased from 7.66±0.34 to 8.47±0.23; Variance value increased from 2.84±1.01E+2 to 9.30±3.53E+2; Correlation value decreased from 1.35±0.45E‐3 to 6.22±2.09E‐4; Cluster shade value increased from 1.20±1.24E+4 to 1.51±1.27E+5; Cluster prominence value increased from 3.11±2.50E+6 to 3.74±3.11E+7. Conclusions: This work has demonstrated the feasibility of ultrasonic texture evaluations of the parotid glands, and the sonographic features may serve as imaging signatures to assess radiation‐induced parotid damage.

SU‐E‐J‐189: The Kullback‐Leiber Divergence for Quantifying Changes in Radiotherapy Treatment Response

PURPOSE Repeated imaging is an extremely powerful tool in current radiotherapy practice since it allows advanced tumor detection and personalized treatment assessment by quantify tumor response. Change detection algorithms have been developed for remote sensing images to mathematically quantify relevant modifications occurring between datasets of the same subject acquired at different times. We propose usage of change detectors in radiotherapy for an automated quantification of clinical changes occurring in repeated imaging. METHODS We explore usage of the Kullback-Leiber divergence as indicator of tumor change and quantification of treatment response. The Kullbach-Leiber divergence uses the likelihood theory to measures the distance between two statistical distributions and thus does not assume consistency in imaging. By its general nature, it can accommodate the presence of noise and variations in imaging acquisition parameters that usually hinder automated identification of clinically-relevant features. RESULTS In a comparison of simple difference maps and the Kullbach-Leiber divergence operator, the difference maps were affected by noise and did not consistently detect changes of low intensity. In contrast, the proposed operator discerned noise by considering regional statistics around each voxel, and marked both regions with low and high contrast changes. CONCLUSIONS Statistical comparison through Kullback-Leiber divergence provides a reliable means to automatically quantify changes in repeated radiotherapy imaging.

SU‐EE‐A2‐03: Evaluation of Auto‐Segmentation Tools for the Target Definition for the Treatment of Lung Cancer

Purpose: With the advent of more sophisticated image devices in the treatment room, image guided radiotherapy(IGRT) and adaptive radiotherapy(ART) have become distinct possibilities. IGRT and ART techniques in their various stages have been implemented in clinics. One of the ART techniques using the daily acquired CTimages involves re‐planning due to the target shape variation during the treatment. Lungcancer volumes of some cases are observed to undergo significant changes where re‐planning is a necessity. To be able to define target efficiently can help the treatment flow significantly. This study evaluates various auto or semi‐auto contouring tools either commercially available or under development for their accuracy and ease of use. Method and Materials: Three methods are included in the study. Two are commercially available (Focal,CMS): auto threshold (of gray level); and auto Segmentation where gray level, the edges and prior shape information are used. The third method is the ITK‐SNAP program that uses a powerful level set(snake) segmentation algorithm to segment anatomical structures in three dimensions. Results: Ten image sets from helical and cone beam CTs are included in the study. The acceptable contours are defined as those with distance to agreement to those drawn by radiation oncologists less than 3 mm. For target volume surrounded by normal lung, the percentage slices of contours that do not need manual adjustment are 41–62%, 23–39%, 62–78% for threshold, auto‐segmentation and SNAP respectively. For cone beam CT, these numbers are approximately 10% lower. SNAP can also be used for target volume with no clear boundary, although the percentage success is much lower. Conclusion: More sophisticated auto‐segmentation tools need to be available routinely with more flexibility for users to adjust algorithm parameters in order for them to be useful for routine clinical ART purposes.

TH-CD-206-02: BEST IN PHYSICS (IMAGING): 3D Prostate Segmentation in MR Images Using Patch-Based Anatomical Signature

PURPOSE MRI has shown promise in identifying prostate tumors with high sensitivity and specificity for the detection of prostate cancer. Accurate segmentation of the prostate plays a key role various tasks: to accurately localize prostate boundaries for biopsy needle placement and radiotherapy, to initialize multi-modal registration algorithms or to obtain the region of interest for computer-aided detection of prostate cancer. However, manual segmentation during biopsy or radiation therapy can be time consuming and subject to inter- and intra-observer variation. This studys purpose it to develop an automated method to address this technical challenge. METHODS We present an automated multi-atlas segmentation for MR prostate segmentation using patch-based label fusion. After an initial preprocessing for all images, all the atlases are non-rigidly registered to a target image. And then, the resulting transformation is used to propagate the anatomical structure labels of the atlas into the space of the target image. The top L similar atlases are further chosen by measuring intensity and structure difference in the region of interest around prostate. Finally, using voxel weighting based on patch-based anatomical signature, the label that the majority of all warped labels predict for each voxel is used for the final segmentation of the target image. RESULTS This segmentation technique was validated with a clinical study of 13 patients. The accuracy of our approach was assessed using the manual segmentation (gold standard). The mean volume Dice Overlap Coefficient was 89.5±2.9% between our and manual segmentation, which indicate that the automatic segmentation method works well and could be used for 3D MRI-guided prostate intervention. CONCLUSION We have developed a new prostate segmentation approach based on the optimal feature learning label fusion framework, demonstrated its clinical feasibility, and validated its accuracy. This segmentation technique could be a useful tool in image-guided interventions for prostate-cancer diagnosis and treatment.

WE‐AB‐207A‐07: A Planning CT‐Guided Scatter Artifact Correction Method for CBCT Images

PURPOSE Cone beam computed tomography (CBCT) imaging is on increasing demand for high-performance image-guided radiotherapy such as online tumor delineation and dose calculation. However, the current CBCT imaging has severe scatter artifacts and its current clinical application is therefore limited to patient setup based mainly on the bony structures. This studys purpose is to develop a CBCT artifact correction method. METHODS The proposed scatter correction method utilizes the planning CT to improve CBCT image quality. First, an image registration is used to match the planning CT with the CBCT to reduce the geometry difference between the two images. Then, the planning CT-based prior information is entered into the Bayesian deconvolution framework to iteratively perform a scatter artifact correction for the CBCT mages. This technique was evaluated using Catphan phantoms with multiple inserts. Contrast-to-noise ratios (CNR) and signal-to-noise ratios (SNR), and the image spatial nonuniformity (ISN) in selected volume of interests (VOIs) were calculated to assess the proposed correction method. RESULTS Post scatter correction, the CNR increased by a factor of 1.96, 3.22, 3.20, 3.46, 3.44, 1.97 and 1.65, and the SNR increased by a factor 1.05, 2.09, 1.71, 3.95, 2.52, 1.54 and 1.84 for the Air, PMP, LDPE, Polystryrene, Acrylic, Delrin and Teflon inserts, respectively. The ISN decreased from 21.1% to 4.7% in the corrected images. All values of CNR, SNR and ISN in the corrected CBCT image were much closer to those in the planning CT images. The results demonstrated that the proposed method reduces the relevant artifacts and recovers CT numbers. CONCLUSION We have developed a novel CBCT artifact correction method based on CT image, and demonstrated that the proposed CT-guided correction method could significantly reduce scatter artifacts and improve the image quality. This method has great potential to correct CBCT images allowing its use in adaptive radiotherapy.

WE‐EF‐210‐06: Ultrasound 2D Strain Measurement of Radiation‐Induced Toxicity: Phantom and Ex Vivo Experiments

Purpose: Radiation-induced fibrosis is a common long-term complication affecting many patients following cancer radiotherapy. Standard clinical assessment of subcutaneous fibrosis is subjective and often limited to visual inspection and palpation. Ultrasound strain imaging describes the compressibility (elasticity) of biological tissues. This study’s purpose is to develop a quantitative ultrasound strain imaging that can consistently and accurately characterize radiation-induce fibrosis. Methods: In this study, we propose a 2D strain imaging method based on deformable image registration. A combined affine and B-spline transformation model is used to calculate the displacement of tissue between pre-stress and post-stress B-mode image sequences. The 2D displacement is estimated through a hybrid image similarity measure metric, which is a combination of the normalized mutual information (NMI) and normalized sum-of-squared-differences (NSSD). And 2D strain is obtained from the gradient of the local displacement. We conducted phantom experiments under various compressions and compared the performance of our proposed method with the standard cross-correlation (CC)- based method using the signal-to-noise (SNR) and contrast-to-noise (CNS) ratios. In addition, we conducted ex-vivo beef muscle experiment to further validate the proposed method. Results: For phantom study, the SNR and CNS values of the proposed method were significantly higher than those calculated from the CC-based method under different strains. The SNR and CNR increased by a factor of 1.9 and 2.7 comparing to the CC-based method. For the ex-vivo experiment, the CC-based method failed to work due to large deformation (6.7%), while our proposed method could accurately detect the stiffness change. Conclusion: We have developed a 2D strain imaging technique based on the deformable image registration, validated its accuracy and feasibility with phantom and ex-vivo data. This 2D ultrasound strain imaging technology may be valuable as physicians try to eliminate radiation-induce fibrosis and improve the therapeutic ratio of cancer radiotherapy. This research is supported in part by DOD PCRP Award W81XWH-13-1-0269, and National Cancer Institute (NCI) Grant CA114313.

TH-E-BRF-09: Gaussian Mixture Model Analysis of Radiation-Induced Parotid-Gland Injury: An Ultrasound Study of Acute and Late Xerostomia in Head-And-Neck Radiotherapy.

PURPOSE Xerostomia (dry mouth), secondary to parotid-gland injury, is a distressing side-effect in head-and-neck radiotherapy (RT). This studys purpose is to develop a novel ultrasound technique to quantitatively evaluate post-RT parotid-gland injury. METHODS Recent ultrasound studies have shown that healthy parotid glands exhibit homogeneous echotexture, whereas post-RT parotid glands are often heterogeneous, with multiple hypoechoic (inflammation) or hyperechoic (fibrosis) regions. We propose to use a Gaussian mixture model to analyze the ultrasonic echo-histogram of the parotid glands. An IRB-approved clinical study was conducted: (1) control-group: 13 healthy-volunteers, served as the control; (2) acutetoxicity group - 20 patients (mean age: 62.5 ± 8.9 years, follow-up: 2.0±0.8 months); and (3) late-toxicity group - 18 patients (mean age: 60.7 ± 7.3 years, follow-up: 20.1±10.4 months). All patients experienced RTOG grade 1 or 2 salivary-gland toxicity. Each participant underwent an ultrasound scan (10 MHz) of the bilateral parotid glands. An echo-intensity histogram was derived for each parotid and a Gaussian mixture model was used to fit the histogram using expectation maximization (EM) algorithm. The quality of the fitting was evaluated with the R-squared value. RESULTS (1) Controlgroup: all parotid glands fitted well with one Gaussian component, with a mean intensity of 79.8±4.9 (R-squared>0.96). (2) Acute-toxicity group: 37 of the 40 post-RT parotid glands fitted well with two Gaussian components, with a mean intensity of 42.9±7.4, 73.3±12.2 (R-squared>0.95). (3) Latetoxicity group: 32 of the 36 post-RT parotid fitted well with 3 Gaussian components, with mean intensities of 49.7±7.6, 77.2±8.7, and 118.6±11.8 (R-squared>0.98). CONCLUSION RT-associated parotid-gland injury is common in head-and-neck RT, but challenging to assess. This work has demonstrated that the Gaussian mixture model of the echo-histogram could quantify acute and late toxicity of the parotid glands. This study provides meaningful preliminary data from future observational and interventional clinical research.