“Object Appearance Models: Construction and Use for Medical Image Segmentation”
Alexandre Xavier Falcão, Universidade de Campinas, Brazil

Segmentation is important to define the spatial extension of body anatomic structures in medical images for quantitative analysis. In this context, it is desirable to eliminate (at least minimize) user interaction in medical image segmentation. This aim is feasible by combining object delineation algorithms with object shape models (OSMs). While the former can better capture the actual shape of the anatomic structures in the image, the latter provides shape constraints to assist structure location and delineation. This lecture will present the construction, use, and comparison between two classes of OSMs for medical image segmentation: statistical (SOSMs) and fuzzy (FOSMs).

SOSMs are very actively pursued and rely on the image mapping into a reference coordinate system, which indicates for each voxel its probability to be in the structure of interest (a probabilistic atlas). Given that the image mapping might not be perfect for a given structure, it is common to complete segmentation with some object delineation algorithm. However, these approaches usually assume that image registration is effective to locate the structure at the right position for delineation. Recent improvements in segmentation accuracy will be shown by using an optimum object search: the atlas translates over the registered image to search for a better object location and delineation, according to some criterion function. Moreover, segmentation methods based on SOSMs tend to use multiple atlases, which considerably increases the processing time to construct and use the models for segmentation. We will show that the optimum object search can reduce the number of atlases with higher segmentation accuracy.

FOSMs avoid the choice of a reference coordinate system and the high computational cost of deformable image registration, by relaxing the shape constraints and relying on the effectiveness of the delineation algorithm used for optimum object search. This considerably reduces the time for model construction. On the other hand, the size of the search region increases, but optimization algorithms and training information obtained during the construction phase can considerably reduce the time for segmentation. As a result, segmentation methods based on FOSMs are significantly faster than methods based on SOSMs, making the former more attractive for studies with a large number of images.

This comparative study will be presented on several human body anatomic structures of the thorax and brain, by using a single OSM per structure. The extension of the work for multiple structures per model is left as future research.

Short CV: Alexandre Xavier Falcão is professor at the Institute of Computing, University of Campinas (UNICAMP), Brazil. He received a B.Sc. in Electrical Engineering from the Federal University of Pernambuco, Brazil, in 1988. He has worked in biomedical image processing, visualization, and analysis since 1991. In 1993, he received a M.Sc. in Electrical Engineering from UNICAMP. During 1994-1996, he worked with the Medical Image Processing Group at the Department of Radiology, University of Pennsylvania, USA, on interactive image segmentation for his doctorate. He got his doctorate in Electrical Engineering from UNICAMP in 1996. In 1997, he worked in a research center (CPqD-TELEBRAS) developing methods for video quality assessment. His experience as professor of Computer Science started in 1998 at UNICAMP. His main research interests include graph algorithms for image processing, model-based and interactive image segmentation, volume visualization, content-based image retrieval, biometry, digital video processing, remote sensing image analysis, active machine learning, pattern recognition, and biomedical image analysis.

“Patient-specific multi-modality visualization environments for minimally invasive image-guided interventions”
Cristian A. Linte, Mayo Clinic, USA

Over the past decades, thanks to the advances in medical image acquisition, visualization and display, surgical tacking and image computing infrastructure, a wide variety of technology has emerged that facilitates diagnosis, procedure planning, intra-operative guidance and treatment monitoring while providing safer and less invasive approaches for therapy delivery. However, while real-time visualization is critical for guidance in absence of direct vision, effective therapy cannot be delivered without the appropriate equipment and instrumentation that enables access to the internal organs through small, less invasive entry routes inside the body. Cardiac interventions have been among the last disciplines to adopt the minimally invasive treatment techniques, mainly due to the challenges associated with access and visualization inside the beating heart.

The lecture will showcase a wide variety of research endeavors on multi-modality imaging environments for cardiac interventions proposed new paradigms in terms of image-guidance technology for beating heart procedures, as well as image-guided interventions and navigation in general. These efforts can be further complemented with the development of more suitable equipment and technology to deliver therapy, while guided by virtual or augmented environments. These technologies include miniature electro-mechanical devices, compatible with traditional imaging modalities, which can be guided inside the body via minimally invasive access ports and remotely manipulated and integrated with commercially available platforms for image-guidance. This lecture will focus both on the technologies (image acquisition, surgical tracking, visualization and display) and techniques (image analysis, modeling, evaluation and validation) currently available and also under development for image-guided (cardiac and not only) interventions, along with their engineering limitations and challenges in translation from bench to bedside. Potential avenues will be highlighted on how to leverage the available infrastructure and expertise in the community, together with the available clinical support and collaborations, to further promote and overcome the slow progress in the clinical translation of image-guided interventions.

Short CV:
Cristian A. Linte holds an academic appointment as Assistant Professor in Biomedical Engineering and Center for Imaging Science at Rochester Institute of Technology in Rochester NY. Dr. Linte received his PhD in Biomedical Engineering at the University of Western Ontario in 2010, under the mentorship of Dr. Terry Peters at Robarts Research Institute. His research focused on the development, evaluation and pre-clinical integration of image guidance techniques for surgical navigation of minimally invasive cardiac interventions. Following his doctoral studies, Cristian spent two years at Mayo Clinic in Rochester, MN at the Biomedical Imaging Resource as a research fellow sponsored by an early career investigator fellowship from the Natural Sciences and Engineering Research Council and the Heart & Stroke Foundation of Canada to investigate novel paradigms for visualization and surgical navigation for minimally invasive therapy.

Dr. Linte's research interests have focused on exploring the use of medical imaging to generate new paradigms for image-guided visualization and navigation for minimally invasive therapy. Dr. Linte's research endeavours have employed both technologies (image acquisition, surgical tracking, visualization and display) and techniques (image analysis, modeling, evaluation and validation) toward the development, evaluation and pre-clinical integration of image guidance environments for surgical navigation of minimally invasive cardiac interventions. His research has been disseminated in more than 60 journal articles and peer-reviewed conference proceedings and has been recognized with several distinctions at international congresses. In 2011, Cristian was recognized with the IEEE MGA GOLD Achievement Award for his leadership and contribution to the IEEE Engineering in Medicine and Biology Society.

At RIT, Dr. Linte established the Biomedical Modeling, Visualization and Image-guided Navigation laboratory dedicated to the discovery and development of innovative imaging, navigation and visualization techniques and instrumentation to improve the understanding, diagnosis and treatment of human diseases through minimally or non-invasive approaches, conducting research across several themes, including medical image processing, cardiac imaging, multi-modality image registration and fusion, modeling and simulation, 3D and stereoscopic visualization, augmented, virtual and mixed reality, instrument tracking, and computer-assisted interventions.

“Image Restoration: A Survey and Recent Advances”
Fiorella Sgallari, University of Bologna, Italy
A real captured image may be distorted by many expected or unexpected factors among which blur and random noise are typical and often unavoidable examples.

Hence, image deblurring and denoising are fundamental tasks in the field of image processing and a plethora of approaches have been proposed throughout the last few decades.

Several image restoration approaches such as nonlinear-diffusion partial differential equations based methods and TV-based regularizations, succeeded in obtaining good quality edge preserving restorations, especially for noise removal. However, they modify images towards piecewise constant functions, in such a way that important information, encoded in image features like textures and details, is often compromised in the restoration process.

We will discuss new adaptive methods for image deblurring and denoising where the regularization operator is constructed by using fractional order derivatives and variational models that uses TV regularization and imposes the resemblance of the residue image to a white noise realization by constraining its autocorrelation function.

Concerning the noise, there are basically three standard noise models in imaging systems: additive noise, multiplicative noise and impulse noise. Typical image noise models are further characterized by the shape of their probability density function, which in the discrete setting is represented by the noise histogram.

In this talk, we will consider first the restoration of images corrupted by additive noise which we assume to be sampled from a known a-priori distribution. Secondly, we discuss a novel variational formulation which exploits the relevant information on the whiteness of the noise thus resulting in a method that is particularly suitable for the restoration of partly-textured perturbed images, since it preserves fine scale features in the restoration process.

Finally, we will introduce the case of image acquired with CCD camera, where the mixture of Poisson and additive-white-Gaussian-noise with blur is a more appropriate model.

New numerical approaches will be discussed and numerical results illustrate the efficiency and effectiveness of different models in image denoising and deblurring.

Short CV:
Fiorella Sgallari is currently professor of Numerical Analysis at the University of Bologna. She is author of more than 120 publications. Her research concerns PDE models and numerical methods for image processing and solution of very large linear systems arising from discrete ill posed problems. She was scientific coordinator of national, international and industrial projects. She organized workshops and international conferences, as Applied Inverse Problems '01 and Scale Space and Variational Methods '07. She is Associate Editor of international journals: Numerical Mathematics: Theory, Methods and Applications, ETNA Electronic Transaction Numer. Analysis, Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, International Journal of Imaging and Robotics and International Journal of Biometrics and Bioinformatics. From 2003 she joined the PhD Committee in Mathematics, University of Bologna. From 2006 to 2013 she was Chair of C. I. R. AM.- Research Centre of Applied Mathematics, University of Bologna.

“High-Fidelity Image-Based Geometric Modeling and Mesh Generation for Engineering Applications”
Yongjie Zhang, Carnegie Mellon University, USA

Image-based geometric modeling and mesh generation play a critical role in computational medicine, biology and engineering. In this talk, I shall present advances and challenges in this area along with a comprehensive computational framework for analysis-suitable geometric modeling and mesh generation, which integrates image processing, geometric modeling, mesh generation and quality improvement with multi-scale analysis at molecular, cellular, tissue and organ scales. The input imaging data are passed through an image-processing module where the image quality is improved. The improved images are then fed to our in-house meshing software, to construct 2D or 3D finite element meshes. Given geometry or atomic resolution data in the Protein Data Bank (PDB), we first construct volumetric density map using a signed distance function or a summation of Gaussian Kernel functions, and then generate various kinds of meshes. Furthermore, the constructed unstructured meshes can be used as control meshes to construct high-order elements such as volumetric T-splines. In addition, a skeleton-based sweeping method is used to generate hexahedral control meshes and solid NURBS (Non-Uniform Rational B-Spline) or cubic Hermite for cardiovascular system. Different from other existing methods, the presented framework supports five important features: multiscale geometric modeling, automatic mesh generation for heterogeneous domains, all-hexahedral mesh generation with sharp feature preservation, robust quality improvement for non-manifold meshes, and high-order element construction. Furthermore, I will also discuss several new advances in volumetric T-spline construction for isogeometric analysis to integrate design and analysis, including feature alignment, trimming curve, conformal surface parameterization and handling extraordinary nodes using subdivision.

Short CV:
Yongjie Jessica Zhang is an Associate Professor in Mechanical Engineering at Carnegie Mellon University with a courtesy appointment in Biomedical Engineering. She received her B.Eng. in Automotive Engineering, and M.Eng. in Engineering Mechanics, all from Tsinghua University, China, and M.Eng. in Aerospace Engineering and Engineering Mechanics, and Ph.D. in Computational Engineering and Sciences from the University of Texas at Austin. Her research interests include computational geometry, mesh generation, computer graphics, visualization, finite element method, isogeometric analysis and their application in computational biomedicine and engineering. She has co-authored over 100 publications in peer-reviewed international journals and conference proceedings. She is the recipient of Presidential Early Career Award for Scientists and Engineers, NSF CAREER Award, Office of Naval Research Young Investigator Award, USACM Gallagher Young Investigator Award, Clarence H. Adamson Career Faculty Fellow in Mechanical Engineering, George Tallman Ladd Research Award, and Donald L. & Rhonda Struminger Faculty Fellow.

"X-ray computed tomography and its applications"
Xiaochuan Pan, The University of Chicago, USA

X-ray computed tomography (CT) remains one of the most widely used tomographic imaging modalities, playing a dominant role in modern medicine and other disciplines such as security scans and material sciences. Since the mid of 1990's, CT has been experiencing a period of renaissance, as a result of the rapid advances in both CT hardware and algorithm developments. The superior spatial/contrast resolution, fast-imaging capability, and high degrees of imaging flexibility offered by modern CT technologies has opened upon ample opportunities for developing innovative applications and imaging protocols in medicine, biology, and material sciences. In the presentation, recent advances in CT systems, algorithm development, image processing, and applications will be discussed, with an emphasis on the introduction of emerging system design of CT imaging for applications in biomedicine, security scan, and material sciences.

Short CV:
Xiaochuan Pan received his Ph.D. degree in Physics at The University of Chicago, followed by the performance of his post-doctoral research in medical imaging, and is a tenured Professor in the Departments of Radiology and Radiation & Cellular Oncology, the College, the Committee on Medical Physics, and the Comprehensive Cancer Center at The University of Chicago. His research interest centers on systems, physics, algorithms, and applications of advanced tomographic imaging. He is a Fellow of AAPM, AIMBE, IAMBE, IEEE, OSA, and SPIE. Awards received by Dr. Pan include IEEE NPSS Early Achievement Award and IEEE EMBS Technical Award for his contributions to advanced medical imaging. Dr. Pan has served as the chair, a charter member, and/or a grant reviewer for review panels of funding agencies and foundations such as NIH, NSF, and NSFC, and is currently an associate editor, or an editorial board member, for a number of journals in the field, such as IEEE Trans. Med. Imaging, IEEE Trans. Biomed Eng., Phys. Biol. Med., Med. Phys., and J. Med. Imaging SPIE. He has served as a chair or member of numerous technical committees of professional organizations such as IEEE AAPM, and RSNA, and as a chair of conference, programs, themes, and sessions, or as a technical or a scientific committee chair or member, for conferences such as IEEE EMBC, IEEE MIC, RSNA, AAPM, and MICCIA in the field.

“Graph Cut, Convex Relaxation and Continuous Max-flow methods for image processing”
Xue-cheng Tai, University of Bergen, Norway

Minimization methods and variational models are becoming fundamental for image processing and computer vision. Graph cut methods, which originated from combinatorial mathematics, have been widely used due to their fast speed and robustness with minimizations. Variational methods are also widely used and they often lead to some complex nonlinear partial differential equations. Fast numerical solvers and robust (global) minimization methods are needed and crucial. Recent research has revealed that graph cut methods (in the discrete setting) and some variational models (in the continuous setting) are solving the same numerical problems. The observation of these connections leads to interesting techniques to convexify some complicated variational models and also to produce fast numerical schemes thanks to some advanced techniques from convex programming.

This tutorial will first introduce graph cut method for image processing and computer vision, then continues with some important variational models. Especially, we will present some recent continuous cut and continuous max-flow models and show their applications to image processing and computer vision. Connection between the discrete graph cut and continuous max-flow models will be revealed. Duality relationship between the different models will be discussed. Convex relaxation of more general variational models will be proposed following these discussions. Fast numerical algorithms becoming natural after convex relaxation and using convex programing techniques. In the end, we will present applications to image segmentation, image restoration, surface construction, machine learning, computer vision and graph theory.

Short CV:
Xue-Cheng Tai received the licentiate degree and the Ph.D. degree in applied mathematics from Jyvaskalya University, Jyvaskalya, Finland in 1989 and 1991, respectively. After holding several research positions in Europe, he became an Associate Professor in 1994 at the University of Bergen, Bergen, Norway, and a Professor in 1997. He was also a faculty member of Nanyang Technological University of Singapore from 2007 to 2010. He has been a Member of the "Center for Mathematics for Applications" in Oslo and a Member of the "Center of Integrated Petroleum Research" in Bergen.

His research interests include Numerical PDEs, optimization techniques, inverse problems, and image processing. He has done significant research work his research areas and published over 100 top quality international conference and journal papers. He is the winner of the 8th Feng Kang Prize for scientific computing. He served as organizing and program committee members for a number of international conferences and has been often invited for international conferences. He has served as referee and reviewers for many premier conferences and journals. Dr. Tai serves as member of the editor board for a number of international journals.