Jinhu Chen

Multiscale registration of medical images based on edge preserving scale space with application in image-guided radiation therapy

Mutual information (MI) is a well-accepted similarity measure for image registration in medical systems. However, MI-based registration faces the challenges of high computational complexity and a high likelihood of being trapped into local optima due to an absence of spatial information. In order to solve these problems, multi-scale frameworks can be used to accelerate registration and improve robustness. Traditional Gaussian pyramid representation is one such technique but it suffers from contour diffusion at coarse levels which may lead to unsatisfactory registration results. In this work, a new multi-scale registration framework called edge preserving multiscale registration (EPMR) was proposed based upon an edge preserving total variation L1 norm (TV-L1) scale space representation. TV-L1 scale space is constructed by selecting edges and contours of images according to their size rather than the intensity values of the image features. This ensures more meaningful spatial information with an EPMR framework for MI-based registration. Furthermore, we design an optimal estimation of the TV-L1 parameter in the EPMR framework by training and minimizing the transformation offset between the registered pairs for automated registration in medical systems. We validated our EPMR method on both simulated mono- and multi-modal medical datasets with ground truth and clinical studies from a combined positron emission tomography/computed tomography (PET/CT) scanner. We compared our registration framework with other traditional registration approaches. Our experimental results demonstrated that our method outperformed other methods in terms of the accuracy and robustness for medical images. EPMR can always achieve a small offset value, which is closer to the ground truth both for mono-modality and multi-modality, and the speed can be increased 5-8% for mono-modality and 10-14% for multi-modality registration under the same condition. Furthermore, clinical application by adaptive gross tumor volume re-contouring for clinical PET/CT image-guided radiation therapy throughout the course of radiotherapy is also studied, and the overlap between the automatically generated contours for the CT image and the contours delineated by the oncologist used for the planning system are on average 90%.

Augmenting atlas-based liver segmentation for radiotherapy treatment planning by incorporating image features proximal to the atlas contours.

Atlas-based segmentation utilizes a library of previously delineated contours of similar cases to facilitate automatic segmentation. The problem, however, remains challenging because of limited information carried by the contours in the library. In this studying, we developed a narrow-shell strategy to enhance the information of each contour in the library and to improve the accuracy of the exiting atlas-based approach. This study presented a new concept of atlas based segmentation method. Instead of using the complete volume of the target organs, only information along the organ contours from the atlas images was used for guiding segmentation of the new image. In setting up an atlas-based library, we included not only the coordinates of contour points, but also the image features adjacent to the contour. In this work, 139 CT images with normal appearing livers collected for radiotherapy treatment planning were used to construct the library. The CT images within the library were first registered to each other using affine registration. The nonlinear narrow shell was generated alongside the object contours of registered images. Matching voxels were selected inside common narrow shell image features of a library case and a new case using a speed-up robust features (SURF) strategy. A deformable registration was then performed using a thin plate splines (TPS) technique. The contour associated with the library case was propagated automatically onto the new image by exploiting the deformation field vectors. The liver contour was finally obtained by employing level set based energy optimization within the narrow shell. The performance of the proposed method was evaluated by comparing quantitatively the auto-segmentation results with that delineated by physicians. A novel atlas-based segmentation technique with inclusion of neighborhood image features through the introduction of a narrow-shell surrounding the target objects was established. Application of the technique to 30 liver cases suggested that the technique was capable to reliably segment liver cases from CT, 4D-CT, and CBCT images with little human interaction. The accuracy and speed of the proposed method are quantitatively validated by comparing automatic segmentation results with the manual delineation results. The Jaccard similarity metric between the automatically generated liver contours obtained by the proposed method and the physician delineated results are on an average 90%-96% for planning images. Incorporation of image features into the library contours improves the currently available atlas-based auto-contouring techniques and provides a clinically practical solution for auto-segmentation. The proposed mountainous narrow shell atlas based method can achieve efficient automatic liver propagation for CT, 4D-CT and CBCT images with following treatment planning and should find widespread application in future treatment planning systems.

A new multiscale registration method for medical image

Mutual information (MI) is a well accepted similarity measure for image registration. However, MI based registration faces the challenges of high computational complexity, low registration efficiency and high likelihood of being trapped into local optima due to an absence of spatial information. In this paper, we propose a new multiscale registration framework based upon an edge preserving total variation L1 norm (TV-L1) scale space representation. Our scale space is constructed by selecting edges and contours of an image according to the geometric size rather than the intensity values of the image features. This ensures more meaningful spatial information for MI based registration. Furthermore, we design an optimal estimation of the TV-L1 parameter in our framework by training and minimizing the transformation offset between the images for automated registration. We validated our method on both simulated mono- and multi-modal medical datasets with ground truth and temporal clinical studies from a combined PET/CT scanner.

A multistep liver segmentation strategy by combining level set based method with texture analysis for CT images

A multi step liver segmentation method is proposed by combining improved level set based method with texture analysis technique for computed tomography (CT) images in this work. The aiming of proposed algorithm is to overcome the segmentation problem which is caused by similar intensities between liver region and its neighboring tissues, also robust to the variations of shape and size within liver region. Firstly, the total variation with the L1 norm (TV-L1) was used for obtaining the initial liver region, which can make the algorithm more efficient and robust. Secondly, both of global and local energy functions with the level set based method are used for extracting the liver region. Finally, the texture analysis method which is based on gray level co-occurrence matrix (GLCM) was used for refining the liver region boundary. The experimental results on 16 clinical planning CT for radiation therapy were used for demonstrating the efficiency of the proposed method both quantitatively and qualitatively.

Parotid gland radiation dose‐xerostomia relationships based on actual delivered dose for nasopharyngeal carcinoma

Abstract Xerostomia induced by radiotherapy is a common toxicity for head and neck carcinoma patients. In this study, the deformable image registration of planning computed tomography (CT) and weekly cone‐beam CT (CBCT) was used to override the Hounsfield unit value of CBCT, and the modified CBCT was introduced to estimate the radiation dose delivered during the course of treatment. Herein, the beams from each patients treatment plan were applied to the modified CBCT to construct the weekly delivered dose. Then, weekly doses were summed together to obtain the accumulated dose. A total of 42 parotid glands (PGs) of 21 nasopharyngeal carcinoma patients were analyzed. Doses delivered to the parotid glands significantly increased compared with the planning doses. V20, V30, V40, Dmean, and D50 increased by 11.3%, 28.6%, 44.4%, 9.5%, and 8.4% respectively. Of the 21 patients included in the study, eight developed xerostomia and the remaining 13 did not. Both planning and delivered PG Dmean for all patients exceeded tolerance (26 Gy). Among the 21 patients, the planning dose and delivered dose of Dmean were 30.6 Gy and 33.6 Gy, respectively, for patients with xerostomia, and 26.3 Gy and 28.0 Gy, respectively, for patients without xerostomia. The D50 of the planning and delivered dose for patients was below tolerance (30 Gy). The results demonstrated that the p‐value of V20, V30, D50, and Dmean difference of the delivery dose between patients with xerostomia and patients without xerostomia was less than 0.05. However, for the planning dose, the significant dosimetric difference between the two groups only existed in D50 and Dmean. Xerostomia is closely related to V20, V30, D50, and Dmean.

SU-F-I-07: CBCT Denoising Based On Adaptive Dictionary Learning Algorithms

PURPOSE We proposed a new dictionary learning algorithm (AK-SVD) based on K-SVD. AK-SVD can denoise the CBCT image, and did not need the noise information as prior knowledge. METHODS The AK-SVD had two steps: signal sparse representation, and then dictionary optimization. The CBCT image was sparse, and there were limited big coefficients. The other coefficients were zero or near zero. In the sparse representation step of traditional K-SVD, the noise variance was used as a threshold to select the big representation coefficients. This increased the complexity of the algorithm. The denoising result also was affected by the accuracy of the noise variance estimation, especially in non-Gaussian noise. In AK-SVD we used the average of the existing big coefficients as a threshold. The new found coefficient was compared with the threshold. If it was bigger than this threshold, it will be determined as the big coefficient, and be added to the set of existing big coefficients. The finding process continued. If it was smaller than this threshold, the finding process was end.This threshold was not related to the noise variance, and based on this method we improved the traditional K-SVD. RESULTS In the synthetic experiments about designing dictionary from synthetic signals, the correct rate of dictionary learning by the AK-SVD was similar with the ideal results of the K-SVD where the noise variance was known. However, the AK-SVD algorithm did not need to evaluate the noise variance, so it had lower computational complexity and wider adaptability. In the denoising experiment about the CBCT image corrupted by the non-Gaussian noise, AK-SVD has an advantage in terms of texture. CONCLUSION The AK-SVD can work well with the noise variance unknown, and it had lower computational complexity and wider adaptability than K-SVD. This work was jointly supported by National Natural Science Foundation of China (61471226), Natural Science Foundation for Distinguished Young Scholars of Shandong Province (JQ201516), China Postdoctoral Science Foundation (2015T80739, 2014M551949), and research funding from Jinan (201401221).

SU‐F‐T‐421: Dosimetry Change During Radiotherapy and Dosimetry Difference for Rigid and Deformed Registration in the Mid‐Thoracic Esophageal Carcinoma

PURPOSE This study aimed to analyze dosimetry changes during radiotherapy for the mid-thoracic esophageal carcinoma, and investigate dosimetry difference between rigid and deformed registration. METHODS Twelve patients with primary middle thoracic esophageal carcinoma were selected randomly. Based on first CT scanning of each patient, plans-o were generated by experience physicists. After 20 fractions treatment, the corresponding plans-re were created with second CT scanning. And then, these two CT images were rigid and deformed registration respectively, and the dose was accumulated plan-o with plan-re. The dosimetry variation of these plans (plan-o: with 30 fractions, plan-rig: the accumulated dose with rigid registration and plan-def: the accumulated dose with deformed registration) were evaluated by paired T-test. RESULTS The V20 value of total lung were 32.68%, 30.3% and 29.71% for plan-o, plan-rig and plan-def respectively. The mean dose of total lung was 17.19 Gy, 16.67 Gy and 16.51 Gy for plan-o plan-rig and plan-def respectively. There were significant differences between plan-o and plan-rig or plan-def for both V20 and mean dose of total lung (with p= 0.003, p= 0.000 for V20 and p=0.008, p= 0.000 for mean dose respectively). There was no significant difference between plan-rig and plan-def (with p=0.118 for V20 and p=0.384 for mean dose). The max dose of spinal-cord was 41.95 Gy, 41.48 Gy and 41.4 Gy for plan-o, plan-rig and plan-def respectively. There were no significant differences for the max dose of spinal-cord between these plans. CONCLUSION The target volume changes and anatomic position displacement of mid-thoracic esophageal carcinoma should not be neglected in clinics. These changes would cause overdose in normal tissue. Therefore, it is necessary to have another CT scanning and re-plan during the mid-thoracic esophageal carcinoma radiotherapy. And the dosimetry difference between rigid and deformed fusions was not found in this study.

SU‐G‐206‐16: Investigation of Dosimetric Consequence Via Cone‐Beam CT Based Dose Reconstruction in Hepatocellular Carcinoma Radiotherapy

PURPOSE Many patients with technically unresectable or medically inoperable hepatocellular carcinoma (HCC) had hepatic dosimetric variations as a result of inter-fraction anatomical deformation. This study was conducted to assess the hepatic dosimetric consequences via reconstructing weekly dose in HCC patients receiving three dimensional conformal radiation therapy. METHODS Twenty-one HCC patients with 21 planning CT (pCT) scans and 63 weekly Cone-beam CT (CBCT) scans were enrolled in this investigation. Among them, six patients had been diagnosed of radiation induced liver disease (RILD) and the other fifteen patients had good prognosis after treatment. And each patient had three weekly CBCT before re-planning. In reconstructing CBCT-based weekly dose, we registered pCT to CBCT to provide the correct Hounsfield units for the CBCT using gradient-based deformable image registration (DIR), and this modified CBCT (mCBCT) were introduced to enable dose calculation.To obtain the weekly dosimetric consequences, the initial plan beam configurations and dose constraints were re-applied to mCBCT for performing dose calculation, and the mCBCT were extrapolated to 25 fractions. Besides, the manually delineated contour was propagated automatically onto the mCBCT of the new patient by exploiting the deformation vectors field, and the reconstructed weekly dose was mapped back to pCT to understand the dose distribution difference. Also, weekly dosimetric variations were compared with the hepatic radiation tolerance in terms of D50 and Dmean. RESULTS Among the twenty-one patients, the three weekly D50 increased by 0.7Gy, 5.1Gy and 6.1Gy, respectively, and Dmean increased by 0.9%, 4.7% and 5.5%, respectively. For patients with RILD, the average values of the third weekly D50 and Dmean were both high than hepatic radiation tolerance, while the values of patients without RILD were below. CONCLUSION The planned dose on pCT was not a real dose to the liver, and the liver overdose increased the risk of RILD. The author would like to express great thanks to Lei Xing, Daniel S Kapp and Yong Yang in the Stanford University School of Medicine for their valuable suggestions to this work. This work is supported by NSFC(61471226), China Postdoctoral Science Foundation (2015T80739,2014M551949) and research funding from Shandong Province (JQ201516).

SU‐F‐J‐221: Adjusted Dose and Its Relation to Radiation Induced Liver Disease During Hepatocellular Carcinoma Radiotherapy

PURPOSE Many patients with hepatocellular carcinoma (HCC) had hepatic anatomy variations as a result of inter-fraction deformation during fractionated radiotherapy, which may result in difference from the planned dose. This study aimed to investigate the relationship between adjusted dose and radiation induced liver disease (RILD) in HCC patients receiving three dimensional conformal radiotherapy (3DCRT). METHODS Twenty-three HCC patients received conventional fractionated 3DCRT were enrolled in this retrospective investigation. Among them, seven patients had been diagnosed of RILD post-radiotherapy, including 4 cases of grade 2, 3 cases of grade 3 according to the CTCAE Version 3.0. Daily cone-beam CT (CBCT) scans were acquired throughout the whole treatment course for each patient. To reconstruct the daily dose to a patient considering the interfraction anatomy variations, the planned beams from each patients treatment plan were firstly applied to each daily modified CBCT (mCBCT). The daily doses were then summed together with the help of deformable image registration (DIR) to obtain the adjusted dose (Dadjusted) of the patient. Finally, the dose changes in normal liver between planned dose (Dplan) and Dadjusted were evaluated by V20, V30, V40 and the mean dose to normal liver (MDTNL). Univariate analysis was performed to identify the significant dose changes. RESULTS Among the twenty-three patients, the adjusted liver V20, V30, V40 and MDTNL showed significant changes from the planned ones (p<0.05) and averagely increased by 4.1%, 4.7%, 4.5% and 3.9Gy, respectively. And the adjusted liver dose in twenty-one patients (91%) were higher than planned value, the adjusted dose of patients with RILD (6/7) exceeds to the hepatic radiation tolerance. CONCLUSION The adjusted dose of all the studied patients significantly differs from planned dose, and mCBCT-based dose reconstruction can aid in evaluating the robustness of the planning solutions, and adjusted dose has the potential to reduce the risk of RILD. The author would like to express great thanks to Lei Xing, Daniel S Kapp and Yong Yang in the Stanford University School of Medicine for their valuable suggestions to this work.This work is supported by NSFC(61471226),China Postdoctoral Science Foundation (2015T80739, 2014M551949) and research funding from Shandong Province(JQ201516).

SU-E-I-100: Heterogeneity Studying for Primary and Lymphoma Tumors by Using Multi-Scale Image Texture Analysis with PET-CT Images

PURPOSE The purpose of this research is studying tumor heterogeneity of the primary and lymphoma by using multi-scale texture analysis with PET-CT images, where the tumor heterogeneity is expressed by texture features. METHODS Datasets were collected from 12 lung cancer patients, and both of primary and lymphoma tumors were detected with all these patients. All patients underwent whole-body 18F-FDG PET/CT scan before treatment. The regions of interest (ROI) of primary and lymphoma tumor were contoured by experienced clinical doctors. Then the ROI of primary and lymphoma tumor is extracted automatically by using Matlab software. According to the geometry size of contour structure, the images of tumor are decomposed by multi-scale method.Wavelet transform was performed on ROI structures within images by L layers sampling, and then wavelet sub-bands which have the same size of the original image are obtained. The number of sub-bands is 3L+1. The gray level co-occurrence matrix (GLCM) is calculated within different sub-bands, thenenergy, inertia, correlation and gray in-homogeneity were extracted from GLCM.Finally, heterogeneity statistical analysis was studied for primary and lymphoma tumor using the texture features. RESULTS Energy, inertia, correlation and gray in-homogeneity are calculated with our experiments for heterogeneity statistical analysis.Energy for primary and lymphomatumor is equal with the same patient, while gray in-homogeneity and inertia of primaryare 2.59595±0.00855, 0.6439±0.0007 respectively. Gray in-homogeneity and inertia of lymphoma are 2.60115±0.00635, 0.64435±0.00055 respectively. The experiments showed that the volume of lymphoma is smaller than primary tumor, but thegray in-homogeneity and inertia were higher than primary tumor with the same patient, and the correlation with lymphoma tumors is zero, while the correlation with primary tumor isslightly strong. CONCLUSION This studying showed that there were effective heterogeneity differences between primary and lymphoma tumor by multi-scale image texture analysis. This work is supported by National Natural Science Foundation of China (No. 61201441), Research Fund for Excellent Young and Middle-aged Scientists of Shandong Province (No. BS2012DX038), Project of Shandong Province Higher Educational Science and Technology Program (No. J12LN23), Jinan youth science and technology star (No.20120109).