ARTIFACTS REDUCTION TECHNIQUES IN X-RAY CONE-BEAM COMPUTED TOMOGRAPHY RECONSTRUCTION A DISSERTATION SUBMITTED TO THE DEPARTMENT OF ELECTRICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Bowen Meng March 2014 © 2014 by Bowen Meng. All Rights Reserved. Re-distributed by Stanford University under license with the author. This dissertation is online at: http://purl.stanford.edu/gf853yf4071 ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Lei Xing, Primary Adviser I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. John Pauly I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Yinyu Ye Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost for Graduate Education This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives. iii iv Abstract CBCT is an important tool in image guided radiation therapy. With the existing re- construction methods, however, various artifacts arise in the image, which are caused by under-sampled projection data, large-area detector, and/or presence of metal im- plants in patients. This dissertation addresses issues for improved CBCT image qual- ity and clinical decision making. In practical CBCT systems, a circular trajectory is commonly used to acquire pro- jections and reconstruct the volume. However, incomplete data might be collected in a short scan either because of mechanical constraint or dose reduction consideration. In such situation, FDK algorithm cannot provide an accurate image due to the theo- reticallimitation. Inthisdissertation, aniterativeoptimizationusingpriorknowledge and rigid image registration is proposed to handle the limited angle reconstruction problem. The algorithm is derived based on the prior image constrained compressed sensing (PICCS) framework. The proposed algorithm is experimentally validated and compared with PICCS algorithm and demonstrates superior reconstruction accuracy. Large-area flat panel detector induces more scatter, which results in cupping and shading artifacts in the images. Various scatter correction methods have been pro- posed to reduce the artifacts from both hardware and software sides, but still suffer clinical applicability. In this dissertation, a single-scan scatter correction method using periphery scatter detection and compressed sensing technique is proposed and tested. The algorithm integrates the scatter measurement/reconstruction and projec- tion acquisition into one scan with simple design of boundary lead blockers. It shows effective scatter artifacts reduction ability as well as promising practical usage for the existing CBCT systems. v The presence of metals in patients may cause streaking artifacts in x-ray CT, which has long been recognized as a problem that not only limits the quality of CT images, but also makes dose calculation in radiation therapy planning problematic. In this dissertation, a method for binary reconstruction of metal objects is proposed to serve as the first step of metal artifacts reduction. The boundaries of metallic objects are obtained by using a penalized weighted least-squares algorithm with the adequateintensitygradient-controlled. Aseriesofexperimentalstudiesareperformed to evaluate the proposed approach, and show that when the projection data are sparse, a non-linear manipulation of projection data can greatly facilitate the binary reconstruction process to achieve accurate binary CT images. vi Acknowledgements First and foremost, I would like to express my sincere gratitude to my advisor, Pro- fessor Lei Xing, for all his generous support, guidance, encouragement and patience during my PhD life at Stanford University. He served as an outstanding researcher and great teacher for me to pursue my studies, for he devoted his precious time and insightful knowledge to me. His mentorship helped me to develop skills of academic thinkings, problem solving as well as creativity and collaboration. I would also like to thank all my oral defense committee members and dissertation committee members for their efforts to review and improve my oral defense and thesis dissertation. I would like to thank Professor Brad Osgood for being my committee chair and offering great help for my oral defense. I would also like to express my grat- itude to Professor John Pauly and Professor Yinyu Ye for their insightful discussion of my thesis and their constructive comments to improve it. I also deeply appreciate Professor Guillem Pratx for giving me many guidance on the defense slides and my research projects. Iwouldalsoliketoacknowledgeafewimportantcollaboratorsandprojectmentors during my PhD study. In particular, I would like to thank Professor Lei Zhu for intro- ducing me to the research group and helping me to start my initial research projects, and Professor Ruijiang Li for giving me many guidances for research projects. I would also like to thanks Professor Jing Wang, Professor Guillem Pratx, Professor Benjamin Fahimian, Professor Yu Kuang for all their insightful discussions for my studies. I would not be able to achieve such progress without their help and discussions. I would also like to thank all the group members at Professor Lei Xing lab for their collaboration, passion, inspiration as well as their constructive discussion and help vii for my research. Their accompanies brought me friendship and enjoyment of the re- search life during these years in addition to that their excellent research achievements provided me many expert advice. I would also like to thank all my friends in United States and China for their sup- portandcareofmystudyandmylife. Theyprovidedmemanyinvaluablesuggestion, help, happiness and unforgettable memories that I will cherish in my life. Lastly and most importantly, I would like to thank my parents and other family members for all their love and encouragement. Without them, I cannot imagine to achieve my PhD today. I hope to express my deepest thanks to my parents for their faith and patience for so many years, as well as guidance for my life. Thanks to their cultivation of my education, personality and attitude that inspired me to complete my PhD study and get ready for the future journey. viii Contents Abstract v Acknowledgements vii 1 Introduction 1 1.1 X-ray Imaging and CT Reconstruction . . . . . . . . . . . . . . . . . 1 1.2 CBCT and FDK Algorithm . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.1 CBCT Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.2 FDK Reconstruction Algorithm . . . . . . . . . . . . . . . . . 6 1.3 CBCT Artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.4 Main Contributions and Publications . . . . . . . . . . . . . . . . . . 9 1.5 Outline of the Dissertation . . . . . . . . . . . . . . . . . . . . . . . . 11 2 Limited Angle Reconstruction for CBCT 13 2.1 Non-Coplanar CBCT Reconstruction . . . . . . . . . . . . . . . . . . 13 2.2 Limited Angle Reconstruction with Prior . . . . . . . . . . . . . . . . 15 2.2.1 Geometry Setup . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.2.2 Existing Algorithms . . . . . . . . . . . . . . . . . . . . . . . 15 2.2.3 Image Registration Facilitated Reconstruction . . . . . . . . . 19 2.2.4 Algorithm Evaluation . . . . . . . . . . . . . . . . . . . . . . . 21 2.3 Reconstruction Results . . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.3.1 Digital Torso Phantom Experiments I . . . . . . . . . . . . . . 22 2.3.2 Digital Torso Phantom Experiments II . . . . . . . . . . . . . 25 2.3.3 Physical Head Phantom Experiments . . . . . . . . . . . . . . 27 ix 2.4 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . 27 3 CBCT Scatter Correction 31 3.1 CBCT Scatter Artifacts . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.2 Scatter Edge Detection and Reconstruction . . . . . . . . . . . . . . . 33 3.2.1 Scatter Correction Using Edge Detection . . . . . . . . . . . . 33 3.2.2 Interpolation Method of Scatter Estimation . . . . . . . . . . 35 3.2.3 Model of Scatter Estimation . . . . . . . . . . . . . . . . . . . 36 3.2.4 Compressed Sensing Optimization . . . . . . . . . . . . . . . . 37 3.2.5 Scatter Correction Procedure . . . . . . . . . . . . . . . . . . 40 3.2.6 Experimental Evaluation . . . . . . . . . . . . . . . . . . . . . 40 3.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3.4 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . 52 4 Binary Reconstruction of Metal Objects 55 4.1 CT Metal Artifacts Reduction . . . . . . . . . . . . . . . . . . . . . . 55 4.2 Binary Reconstruction of Metal Objects . . . . . . . . . . . . . . . . 57 4.2.1 Pre-processing of Projection Data . . . . . . . . . . . . . . . . 57 4.2.2 Gradient-Controlled PWLS algorithm . . . . . . . . . . . . . . 58 4.2.3 Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . 60 4.3 Binary Reconstruction Results . . . . . . . . . . . . . . . . . . . . . . 62 4.3.1 Head Phantom with Fiducial Markers . . . . . . . . . . . . . . 62 4.3.2 Head Phantom with Triangularly Brass . . . . . . . . . . . . . 67 4.3.3 Head Phantom with Metal Nut . . . . . . . . . . . . . . . . . 69 4.4 Discussion and Conclusions . . . . . . . . . . . . . . . . . . . . . . . 73 5 Summary and Future Research 77 A ADMM Algorithm in Section 3.2.4 79 B Glossary of Symbols Used in Chapter 1, 2, 3 and 4 81 Bibliography 83 x