Preface
The development of medical imaging brought new challenges to nearly all of the various fields of image processing. Current solutions for image registration and pattern matching need to deal with multiple image modalities since a diagnostic of a patient may need several different kind of imagery (Ferrant 2001, Ferrant 2002, Ino 2003, Samant 2008, Warfield 1998, Warfield 2002). Image registration is also an important part of image guided surgery systems (Grimson 1996, Sugano 2001). Automated recognition and diagnosis require image segmentation, quantification and enhancement tools (BachCuadra 2004, Davatzikos 1995, Freedman 2005, Mcinerney 1996, Savelonas 2009, Wareld 1995, Wells 1996). Particularly, image segmentation and tracking are crucial to cope with the increasing amount of data encountered in medical applications and became crucial for medical image analysis and classification (Hadjiiski 1999, Park 1996, Sahiner 2001). Generation of complex visualizations, like internal organs and brain, became essential in disease diagnosis or patient care. Three-dimensional visualization has proven to be a valuable support tool, especially for the assessment of the shape and position of anatomical structures. A spatial structure is generally easy to comprehend in such view than in a series of cross sections; thus some cognitive load can be taken off the viewer. 3D image reconstruction applications can be used for reporting a diagnostic of a patient avoiding, at least in a first phase, the step to surgery or to provide cross-section slices with information about the tissues contained in the considered slices (Braumann 2005, Chen 1990, Norton 2003, Yagel 1997). In medical imagery, the resolution of 3D representations needs to be high, in order to get the maximum of geometrical details. However such detailed surfaces can be a major drawback for an efficient use of such data. They imply the archival or storage of a large quantity of similar data in a patient database and the communication of large amounts of data during clinical diagnosis or follow-up cares of patients, only to mention the most trivial implications. These facts justify the development of efficient and effective compression methods dedicated to medical imagery. Consequently biomedical diagnostic has been, and still is, greatly improved with all the advances on medical imaging.
The acquisition devices constantly advance towards providing more information in the images, improved resolutions, better repeatability and quality of the acquisitions and faster acquisition. Also the algorithms operating over those images have showed an increasing ability of extracting the information from incomplete and damaged data and incorporate prior knowledge into each acquisition. However, with the constant improvement of the medical scanners, some new challenges have been created for the image processing algorithms.
The images from those devices started to show ever improving quality, resolution and they have been acquired in shorter times. No longer was it necessary to incorporate computational power only towards tasks like denoising or improvement of incomplete data. The amount of data delivered by the medical devices started to grow so significantly that it was necessary to provide more and more computational power in order to properly analyze all the available information in reasonable times. The new interest in medical image processing towards fully 3D processing, which targets improving the information retrieval from the raw image material, also brings new challenges. Three dimensional image acquisition devices like computer tomography (CT), magnetic resonance imaging (MRI), or 3D-ultrasound (3D-US) are increasingly used to support and facilitate medical diagnosis. 3D image processing is a key technique for preoperative surgical planning and intraoperative guidance systems (DiGioia 1998, Ellis 1999, Kawasaki 2004, Kikinis 1996, Liu 2007, Ma 2003). Processing in 3D can capture more of the information in the image data, which can improve the attainable quality of the results. The large amounts of data and the complexity of the 3D methods, obviously, imply long processing times.
The quantity of information started to grow more rapidly than the modern processing units could have keep up with. This has been aligned in time with the turn in the approach for development of modern microprocessors. Until recently we have been witnessing the growing capabilities of new microprocessors in very notable and predictable manner. Their speed has been increasing constantly with the respect to Moore’s law, which meant that every algorithm designed at that time would perform better if just executed on more modern processing unit. This trend has changed recently and the producers started to expand the possibilities of their microprocessors by equipping them with several cores and thus giving them the capabilities of parallel processing. Therefore, the concept of parallel processing has migrated from advanced computing centers towards any personal computer (Foster 1998). This fact, together with the growing sizes of medical data, has attracted numerous researchers towards the application of High Performance Computing (HPC) to medical imaging data. Several have been presented above, but others can be found in literature (Kawasaki 2003, Kikinis 1998, Lenkiewicz 2009, Liao 2002, Mizuno-Matsumoto 2000, Ourselin 2002, Rohlfing 2003, Thompson 1997, Warfield 2002).
The most common representatives of the HPC technology have been mainly the computer clusters. A computer cluster is a set of homogeneous computers usually connected through fast local area networks and working together with some network services. Clusters are commonly used to improve performance and/or availability in a cost-effective manner regarding a sole computer with similar computing power or availability (Barbosa 2000, Papadonikolakis 2008). Grid computing goes beyond cluster computing in the sense that nodes can be heterogeneous in terms of hardware and software can be loosely connected across subnets or domains and can be dynamic in terms of the number of nodes that can enter or leave the grid over time. There are several applications for both cluster and grid computing, including biomedical applications (Ataseven 2008, Benkner 2003, Liakos 2005, Mayer 1999). On the other hand, there is an increasing interest on the application of computing technologies to advance biomedical diagnostics and clinical technologies. Computer assisted diagnosis and therapy plays nowadays a key role to decrease young or mid-age mortality, to improve quality of live or to increase life expectancy in aging societies.
In the last three decades computer clusters have served for numerous research centers, government institutions and private enterprises. The technology was rather limited for those institutions as it was expensive and difficult to utilize. The recent progress has turned that trend, as it became relatively inexpensive to obtain a multi-processing unit platform for entities like universities and small enterprises, or virtually any household. This fact of highly increased availability has started a growth in the number of solutions created for this field, namely operating systems, management software and programming tools.
Until recently, medical image analysis has not been a traditional field of application for high performance computing. Furthermore, the amount of data produced by the different scanners has been relatively moderate. Several developments are now moving the field of medical image analysis towards the use of high performance computing techniques. Specifically, this is because of the increase of computational requirements, of the data volume and of the intensity of electronic utilization of the data. It is necessary to further validate and test the robustness of the algorithms by applying them to a larger number of cases, normally to medical databases. The huge volume of data generated by some medical and biological applications may require special processing resources, while guaranteeing privacy and security. Cluster and grid computing become crucial in dealing with huge sensitive data for special medical activities, such as diagnosis, therapy, e-health, tele-surgery, as well as specific domains, such as phylogenetics, genomics and proteomics, or studying virus evolution and molecular epidemiology.
This book explores synergies between biomedical diagnostics and clinical technologies and high-performance cluster and grid computing technologies addressing selected topics in these areas. Special attention is paid to biomedical diagnostics and clinical technologies with critical requirements in terms of processing power, namely 2D/3D medical image segmentation and reconstruction, volumetric texture analysis in biomedical imaging, modeling and simulation in medicine, computer aided diagnosis, analysis of doppler embolic signals, and massive data classification of neural responses.
ORGANIZATION OF THE BOOK
The book is organized into nine chapters focusing some of most important challenges previously presented, namely segmentation, tracking, registration, reconstruction, visualization, compression, classification, and analysis of medical imagery. They are examples on how high performance computing is applied to medical data and bioinformatics in each specific topic. Two Selected readings were joined for additional reading to complete the book with two important topics: the co-registration and the fusion.
A brief description of each of the chapters follows:
The first chapter provides an overview about most common techniques for 2D/3D medical image segmentation. Most important features, examples of the results that they can bring and examples of application are presented. Promising recent methods are evaluated and compared based on a selection of important features. This chapter is not only an overview but also serves to show which are the directions currently taken by researchers and which of them have the potential to be successful.
The second chapter focuses on 2D/3D medical image segmentation and tracking through statistical region-based active contour models where the region descriptor is selected as the probability density function of an image feature. They focus on active contours or surfaces that are particularly well adapted to the treatment of medical structures since they provide a compact and analytical representation of object shape. Successful applications of this model are described for brain magnetic resonance imaging (MRI), contrast echocardiography, and perfusion MRI.
The third chapter explores modeling and simulation of deep brain stimulation (DBS) in Parkinson disease and intends to explain how models of neurons and connected brain nuclei contribute to the understanding of deep brain stimulation. For that purpose a selection of models capable of describing one or more of the symptoms of Parkinson’s disease are presented. A discussion of how models of neurons and connected brain nuclei contribute to the understanding of DBS is presented.
The fourth chapter addresses the problem of high-performance 3D image reconstruction for medical imaging, based on anatomical knowledge, in a time-frame compatible with the workflow in hospitals. The chapter presents an overview of reconstruction algorithms applicable to medical computed tomography and high-performance 3D image reconstruction. Different families of modern HPC platforms and the optimization methods that are applicable for the different platforms are also discussed in this chapter.
The fifth chapter presents a lossy compression algorithm for triangular meshes that may be used in medical imaging. The presented scheme is based on wavelet filtering, and an original bit allocation process that optimizes the quantization of the data. Authors demonstrate that an efficient allocation process allows the achievement of good compression results for a very low computational cost but also present similar visual results when compared to a lossless coder at medium bitrates.
The sixth chapter presents self-similarity models that can be applied to Breast Cancer detection through Digital Mammography. Human tissue is characterized by a high degree of self-similarity, and that property has been found in medical images of breasts, through a qualitative appreciation of the existing self-similarity nature, by analyzing their fluctuations at different resolutions. Authors conclude that self-similarity measure can be an excellent aid to evaluate cancer features, giving an indication to the radiologist diagnosis.
The seventh chapter presents an overview on volumetric texture analysis for medical imaging. Focus was done to the most important texture analysis methodologies, those that can be used to generate a volumetric Measurement Space. The methodologies presented are: Spatial Domain techniques, Wavelets, Co-occurrence Matrix, Frequency Filtering, Local Binary Patterns and Texture Spectra and The Trace Transform. The chapter ends with a review of the advantages and disadvantages of the techniques and their current applications and present references to where the techniques have been used.
The authors conclude that texture analysis presents an attractive route to analyze medical or biological images and will play an important role in the discrimination and analysis of biomedical imaging.
The eighth chapter describes an integrated view of analysis of Doppler embolic signals using high-performance computing. Fundamental issues that will constrain the analysis of embolic signals are addressed. Major diagnostic approaches to Doppler embolic signals focuses on the most significant methods and techniques used to detect and classify embolic events including the clinical relevancy are presented. The survey includes the main domains of signal representation: time, time-frequency, time-scale and displacement-frequency.
The ninth chapter proposes and assesses experimentally a platform for the mass-classification of neuronal responses using data-parallelism for speeding up the classification of neuronal responses. Authors conclude that a significant computational speed-up can be achieved by exploiting data level parallelism for the classification of the neural response in each electrode.
The next two chapters are two Selected Readings.
The first one is devoted to the development of a framework for image and geometry co-registration by extending the functionality of the widely used Visualization Toolkit (VTK). A real application in stereotactic radiotherapy treatment planning that is based on the particular framework is presented.
The second one is devoted to biomedical image registration and fusion, in order to assist medical knowledge discovery by integrating and simultaneously representing relevant information from diverse imaging resources. An overview on fundamental knowledge and major methodologies of biomedical image registration, and major applications of image registration in biomedicine is presented.
The purpose of this book is to provide a written compilation for the dissemination of knowledge and to improve our understanding about high performance computing in biomedical applications. The book is directed for Master of Science or Doctor of Philosophy students, researchers and professionals working in the broad field of biomedical informatics. We expect that the deep analyses provided inside the book will be valuable to researchers and others professionals interested in the latest knowledge in these fields.
REFERENCES
Ataseven, Y., Akalin-Acar, Z., Acar, C., & Gençer, N. (2008). Parallel implementation of the accelerated BEM approach for EMSI of the human brain. Medical and Biological Engineering and Computing, 46(7), 671-679.
BachCuadra, M. (2004). Atlas-Based Segmentation of Pathological MR Brain Images Using a Model of Lesion Growth. IEEE Transactions on Medical Imaging 23(10), 1301- 1314.
Barbosa, J., Tavares, J., & Padilha, A. (2001). Parallel Image Processing System on a Cluster of Personal Computers. Vector and Parallel Processing (pp. 439-452).
Benkner, S., Dimitrov, A., Engelbrecht, G., Schmidt, R., & Terziev, N. (2003). A Service-Oriented Framework for Parallel Medical Image Reconstruction. Computational Science, 691-691.
Braumann, U., Kuska, J., Einenkel, J., Horn, L., Luffler, M., & Huckel, M. (2005). Three- dimensional reconstruction and quantification of cervical carcinoma invasion fronts from histological serial sections. IEEE Transactions on Medical Imaging, 24(10), 1286-1307.
Chen, C., Lee, S., & Cho, Z. (1990). A parallel implementation of 3d CT image reconstruction on a hypercube multiprocessor. IEEE Transactions on Nuclear Science, 37(3), 1333--1346.
Davatzikos, C., & Prince, J. L. (1995). An Active Contour Model for Mapping the Cortex. IEEE Transactions on Medical Imaging, (14), 65-80.
DiGioia, A. M., Jaramaz, B., Blackwell, M., Simon, D. A., Morgan, F., Moody, J. E., Nikou, C., Colgan, B. D., Astion, C. A., Labarca, R. S., Kischell, E., & Kanade, T. (1998). Image guided navigation system to measure intraoperatively acetabular implant alignment. Clinical Orthopaedics and Related Research, (355), 8-22.
Ellis, R. E., Tso, C. Y., Rudan, J. F., & Harrison, M.M. (1999). A surgical planning and guidance system for high tibial osteotomy. Computer Aided Surgery, 4(5), 264-274.
Ferrant, M., Nabavi, A., Macq, B., Black, P. M., Jolesz, F. A., Kikinis, R., & Warfield, S. K. (2002). Serial registration of intraoperative MR images of the brain. Medical Image Analysis, 6(4), 337-359.
Ferrant, M., Nabavi, A., Macq, B., Jolesz, F. A., Kikinis, R., & Warfield, S. K. (2001). Registration of 3-D intraoperative MR images of the brain using a finite-element biomechanical model. IEEE Transaction on Medical Imaging, 20(12), 1384-1397.
Foster, I. & Kesselman, C. (Ed.). (1998). The Grid: Blueprint of a New Computing Infrastructure. San Mateo, CA: Morgan Kaufmann Publishers.
Freedman, D., Radke, R. J., Zhang, T., Jeong, Y., Lovelock, D. M., & Chen, G. T. Y. (2005). Model-based segmentation of medical imagery by matching distributions. IEEE Transactions on Medical Imaging, 24(3), 281-292.
Grimson, W., Ettinger, G., White, S., Lozano-Perez, T., Wells, W., & Kikinis, R. (1996). An, Automatic Registration Method for Frameless Stereotaxy, Image Guided Surgery, and, Enhanced Reality Visualization. IEEE Transactions On Medical Imaging, (15), 129-140.
Hadjiiski, L., Sahiner, B., Chan, H.P., Petrick, N., & Helvie, M. (1999). Classification of malignant and benign masses based on hybrid ART2LDA approach. IEEE transactions on medical imaging 18(12), 1178-87.
Ino, F., Ooyama, K., Kawasaki, Y., Takeuchi, A., Mizutani, Y., Masumoto, J., Sato, Y., Sugano, N., Nishii, T., Miki, H., Yoshikawa, H., Yonenobu, K., Tamura, S., Ochi, T., & Hagihara, K. (2003). A high performance computing service over the Internet for nonrigid image registration. In Proceedings on Computer Assisted Radiology and Surgery: 17th International Congress and Exhibition (pp. 193-199).
Kawasaki, Y., Ino, F., Mizutani, Y, Fujimoto, N., Sasama, T., Sato, Y., Tamura, S., & Hagihara, K. (2003). A High Performance Computing System for Medical Imaging in the Remote Operating Room. High Performance Computing (LNCS 2913, pp. 162-173). Berlin/Heidelberg: Springer.
Kawasaki, Y., Ino, F., Mizutani, Y., Fujimoto, N., Sasama, T., Sato, Y., Sugano, N., Tamura, S., & Hagihara, K. (2004). High-Performance Computing Service Over, the Internet for Intraoperative Image Processing. IEEE Transactions On Information Technology In Biomedicine, 8(1), 36-46.
Kikinis, R., Shenton, M. E., Iosifescu, D. V., McCarley, R. W., Saiviroonporn, P, Hokama, H. H., Robatino, A., Metcalf, D., Wible, C. G., Portas, C. M., Donnino, R. M., & Jolesz, F. A. (1996). A Digital Brain Atlas for Surgical Planning, Model Driven Segmentation, and Teaching. IEEE Transactions on Visualization and Computer Graphics, (2), 232-241.
Kikinis, R., Warfield, S. K., & Westin, C. F. (1998). High Performance Computing in Medical Image Analysis at the Surgical Planning Laboratory. High Performance Computing, 290-297.
Lenkiewicz, P., Pereira, M. Freire, M. & Fernandes, J. (2009). A New 3D Image Segmentation Method for Parallel Architectures. In Proceedings of 2009 IEEE International Conference on Multimedia and Expo (ICME 2009), New York, USA, June 28 - July 3, 2009 (pp. 1813-1816). IEEE Press.
Liakos, K., Burger, A., & Baldock, R. (2005). Distributed Processing of Large BioMedical 3D Images. High Performance Computing for Computational Science,142-155.
Liao, H., Hata, N., Iwahara, M., Nakajima, S., Sakuma, I., & Dohi, T. (2002). High-resolution stereoscopic surgical display using parallel integral videography and multi-projector. In Proceedings of 5th International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 85-92).
Liu, W., Schmidt, B., Voss, G, & Muller-Wittig, W. (2007). Streaming Algorithms for Biological Sequence Alignment on GPUs. IEEE Transactions on Parallel and Distributed Systems, 18(9), 1270-1281.
Ma, B., & Ellis, R. E. (2003). Robust registration for computer-integrated orthopedic surgery: Laboratory validation and clinical experience. Medical Image Analysis, 7(3), 237-250.
Mayer, A., & Meinzer, H. P. (1999). High performance medical image processing in client/server-environments. Computer Methods and Programs in Biomedicine, 58, 207-217.
Mcinerney, T., & Terzopoulos, D. (1996). Deformable Models in Medical Image Analysis: A Survey. Medical Image Analysis, (1), 91-108.
Mizuno-Matsumoto, Y., Date, S., Tabuchi, Y., Tamura, S., Sato, Y., Zoroofi, R. A., Shimojo, S., Kadobayashi, Y., Tatsumi, H., Nogawa, H., Shinosaki, K., Takeda, M., Inouye, T., & Miyahara, H. (2000). Telemedicine for evaluation of brain function by a metacomputer. IEEE Transaction on Information. Technology. Biomedicine, 4(2), 65-172.
Norton, A. & Rockwood, A. (2003). Enabling view-dependent progressive volume visualization on the grid. IEEE Computer Graphics and Applications, 23(2), 22-31.
Ourselin S., Stefanescu R., & Pennec X. (2002). Robust registration of multimodal images: Towards real-time clinical applications. In Proceedings of 5th International Conference Medical Image Computing and Computer-Assisted Intervention (pp. 140-147).
Papadonikolakis, M., Kakarountas, A., & Goutis, C. (2008). Efficient high-performance implementation of JPEG-LS encoder. Journal of Real-Time Image Processing, 3(4), 303-310.
Park, J., Metaxas, D., Young, A. A., & Axel, L. (1996). Deformable Models with Parameter Functions for Cardiac Motion Analysis from Tagged MRI Data. IEEE Transactions on Medical Imaging, 437-442.
Rohlfing, T. & Maurer, C. R. (2003). Nonrigid image registration in shared memory multiprocessor environments with application to brains, breasts, and bees. IEEE Transaction on Information. Technology in Biomedicine, 7(1), 16-25.
Sahiner, B., Petrick, N., Chan, H. P., Hadjiiski, L. M., Paramagul, C., Helvie, M. A., & Gurcan, M. N. (2001). Computer-aided characterization of mammographic masses: accuracy of mass segmentation and its effects on characterization. IEEE transactions on medical imaging, 20(12), 1275-84.
Samant, S. S., Xia, J., Muyan-Özçelik P. & Owens J. D. (2008). High performance computing for deformable image registration: Towards a new paradigm in adaptive radiotherapy. Medical Physics, 35(8), 3546-3553.
Savelonas, M.A., Iakovidis D. K., Legakis I., & Maroulis, D. (2009). Active Contours Guided by Echogenicity and Texture for Delineation of Thyroid Nodules in Ultrasound Images. IEEE Transactions on Information Technology in Biomedicine, 13(4), 519-527.
Sugano, N., Sasama, T., Sato, Y., Nakajima, Y., Nishii, T., Yonenobu, K., Tamura, S., & Ochi, T. (2001). Accuracy evaluation of surface-based registration methods in a computer navigation system for hip surgery performed through a posterolateral approach. Computer Aided Surgery, 6(4), 195-203.
Thompson, P. M., MacDonald, D., Mega, M. S., Holmes, C. J., Evans, A. C., & Toga, A. W. (1997). Detection and Mapping of Abnormal Brain Structures with a Probabilistic Atlas of Cortical Surfaces. Journal of Computer Assisted Tomography, 21(4), 567-581.
Wareld, S., Dengler, J., Zaers, J., Guttmann, C. R., Wells, W. M., Ettinger, G. J., Hiller, J., & Kikinis, R. (1995). Automatic identication of Grey Matter Structures from MRI to Improve the Segmentation of White Matter Lesions. Journal of Image Guided Surgery, 6(1), 326-338.
Warfield, S. K., Jolesz, F. A., & Kikinis, R. (1998). A high performance computing approach to the registration of medical imaging data. Parallel Computing, 24(9-10), 1345-1368.
Warfield, S. K., Talos, F., Tei, A., Bharatha, A., Nabavi, A., Ferrant, M., Black, P. M., Jolesz, F. A., & Kikinis, R. (2002). Real-time registration of volumetric brain MRI by biomechanical simulation of deformation during image guided surgery. Computing and Visualization in Science, 5(1), 3-11.
Wells, W. M.,. Grimson, W.E.L., Kikinis, R., & Jolesz, F. A. (1996). Adaptive Segmentation of MRI Data. IEEE Transactions on Medical Imaging 15(4), 429-442.
Yagel, R., Mueller, K., Fredrick Cornhill, J. & Mueller, K. (1997). The Weighted Distance Scheme: A Globally Optimizing Projection Ordering Method for ART. IEEE Transactions on Medical Imaging, (16), 223-230.