JOURNAL OF COMPUTER SCIENCE AND ENGINEERING, VOLUME 2, ISSUE 2 AUGUST 2010 1

An Image Knowledge and InformationSharing System for a Biomedical Informatics database to Promote Translational Research Teeradache Viangteeravat, Matthew N. Anyanwu, Venkateswara Ra Nagisetty, Emin Kuscu , Ian M. Brooks and Chanchai S. McDonald

Abstract— Large datasets are increasingly becoming a hallmark of basic science and clinical research. The heterogeneous nature of these datasets can mean that both text and image data of various modalities must be combined into a coherent whole. To provide for diverse image-data management needs in clinical and basic science we have developed image-processing tools mounted on a digital management system. We named the system “Image Knowledge and Information Sharing System” (IKISS). In addition to image management, IKISS has functionalities for knowledge exchange and online collaboration. IKISS is accessed via user-friendly web-based applications, thus increasing image accessibility and encouraging collaboration between researchers and/or clinicians, an essential factor in the development of translational research. IKISS is a model for how using a central image management system can improve efficiency of collarborations, and bridge the gap between laboratory discovery and clinical practice.

Index Terms— Basic Research, Bioinformatics, Clinical Trial, Data Sharing, Health Management, Image Sharing, Laboratory Management.

——————————  ——————————

1 INTRODUCTION

R

APID advances and improvements in Internet network speed and browser capability in recent years have allowed the creation of increasingly sophisticated suites of online tools and applications to support scientific research across disciplines and research areas. Many web-based data management systems have been developed, offering the potential to bridge research silos and provide a format for real-time inter- and intra-group collaborations. Because data and knowledge can now be more easily shared between research domains, this expansion of browsing capabilities and increased network performance has helped increase opportunities for ‘translational research’. Translational research applies discoveries from basic science laboratories to clinical care situations, either as Type I (bench-to-bedside) or Type II (bedside to community) research. Such efforts combine the skills of the basic laboratory researcher and the clinician

into patient-centered science and on a larger scale, clinical interventions may be translated into community initiatives and public health policy. It is possible for translational research clinician-scientists to analyze, interpret and even possibly diagnose from networked image data. Given this expansion of translational science technologies, a basic component to be considered when developing centralized integrated data management systems is the ability to allow data collection and knowledge extraction from an image, and its augmentation with pertinent experiential input. To address such issues, the “Image Knowledge and Information Sharing System” (IKISS) was developed by the Biomedical Informatics Unit (BMIU) at the University of Tennessee Health Science Center (UTHSC). IKISS was developed in response to user requests for tools to aid in collaborative sharing of images, and accumulation and integration of different image modalities in both the basic ———————————————— sciences and clinical research. IKISS is a plug-in module T. Viangteeravat is with the Clinical and translation Institute, University for the Scientific Laboratory Information Managementof Tennessee Health Science Center. M.N. Anyanwu is with the Clinical and translation Institute, University of Patient-care Research Information Management (SlimTennessee Health Science Center. Prim) system, developed and deployed by BMIU [1, 2]. V. Nagisetty is with the Clinical and translation Institute, University of Tennessee Health Science Center. E. Kuscu is with the Clinical and translation Institute, University of Tennessee Health Science Center. I.M. Brooks is with the Clinical and translation Institute, University of Tennessee Health Science Center. C.S. McDonald is with the, University of Tennessee Health Science Center.

© 2010 JCSE http://sites.google.com/site/jcseuk/

JOURNAL OF COMPUTER SCIENCE AND ENGINEERING, VOLUME 2, ISSUE 2 AUGUST 2010 2

Fig. 1. Image Interraction feature:side-by-side comparison

2 BACKGROUND Various data management techniques, with varying levels of complexity, are available to researchers [e.g. 1, 2, 3, 4, 7, 8]. An early example of medical image viewing and report generating was the Java-based remote viewing station (JaRViS) [6]. JaRViS exploited local-area network systems for web-based image processing of diagnostic images generated through nuclear medicine. Picture Archiving and Communication Systems (PACS) and was introduced as a primary image processing tool for digital image management for web-based image processing of diagnostic images generated through nuclear medicine. Image data were housed on a central server and accessed through secure web service [6] The PACS system architecture works well as a local area network [4, 7]. However because the system is deployed on a secure hospital server it is not an effective tool for intra-group communication, nor can data be searched and downloaded by external collaborators, such as laboratory scientists working at separate locations. Kim et al. proposed the Functional Imaging Web (FIWeb) [7]. The FIWeb is a web-based medical image data processing and management system. The FIWeb system uses Python and JavaScript for rendering a graphical user interface (GUI) and uses Java Applets for development of online image processing functions [7]. Kalinski et. al., introduced virtual 3D microscopy using JPEG200 for the visualization of pathology specimens in Digital Imaging and Communications in Medicine (DICOM) format to create knowledge databases and online learning platforms [8, 9]. Recently, Harris et al. proposed the Research Electronic Data Capture (REDCap) project [3]. REDCap is used for clinical trial management and uses PHP and JavaScript programming languages [3, 5, 10]. The REDCap system uses a flat-table structure on a MySQL server for collecting and storing research data. Although this allows for efficient study setup and data export, it is not ideal for creating data repositories [3].

Importantly, image modality functions are not yet available in REDCap. To address some of the issues in current data repository systems we recently introduced a biomedical informatics data management system called Slim-Prim (Scientific Laboratory Information Management-Patient-care Research Information Management) [1, 2]. Slim-Prim is a data repository for basic and clinical science data, written in PHP and JavaScript [5] and mounted on an Oracle 10g database engine, which is a relational database management system.This makes it possible to relate data items within the databasethus making it easier to search for two or more related items in the database based on the input from the user. Although Slim-Prim was launched with functionalities to address complex data integration issues using ontology-based data-mapping concepts, at first release an image knowledge-processing tool was not available in the system. Finally, several Centers and Institutes of the National Institutes of Health (NIH) have invested heavily in individual image-data storage and retrieval systems. Notable examples are the National Center for Research Resources “Biomedical Informatics Research Network” (BIRN) and the National Cancer Institute “cancer Biomedical Informatics Grid” (caBIG). The BIRN system [11] extracts/retrieves and then transmits images from a source, while caBIG manages oncology and radiology images from multiple sources through its web servers [12]. Importantly, both these systems encourage and foster collaboration between individuals and research groups. We believe the BMIU IKISS system adds to the functionality of these endeavors. IKISS combines the functionalities of Slim-Prim clinical data management (CDM) features with non-clinical image management features, thus making IKISS a novel approach in CDM.

Fig. 2. IKISS API to retrieve sagittal images(DAB2) from Allen Institute for Brain Science

JOURNAL OF COMPUTER SCIENCE AND ENGINEERING, VOLUME 2, ISSUE 2 AUGUST 2010 3

3 IKISS SYSTEM 3.1 IKISS ARCHITECTURE AND ALGORITHM The IKISS system architecture comprises four main components, which provide an interface to the MATLAB signal processing and image processing toolboxes. 1. A web based application that uses PHP and JavaScript programming languages with inverted file indexing search algorithms to render the graphical user interface (GUI) [5]; 2.

A MySQL database engine for data storage [10];

3.

An image analysis tool composed of seven subcomponents (explained further below - Region of Interest (ROI) tooltip, Catalog, Comment, Highlight, Chat, Pan View, PubMed Reference);

4.

An image processing tool comprising three subcomponents: convert, crop, and rotation.

The system architecture is extensible and customizable, such that new components can be added to the system without affecting other components. IKISS architecture is similar to the “Pixelpost” content management system (CMS) [13] that uses a catalogmodeling structure.. The CMS retrieves images from the database using an inverted file indexing structure. This is faster and more efficient than other indexing structures/schemes such as hash tables or index trees. The CMS presents a side-by-side comparison of two images for users to compare features between like images, for example histological brain slices (Figure 1). The algorithm also provides metadata for the images. The high resolution and high quality images produced are vertical, i.e., domain specific (from one data source) as opposed to images produced by other, traditional web-based search engines, which usually originate from two or more domains [14]

3.2 IKISS APPLICATION PROGRAMMING INTERFACE The Application Programming Interface (API) of the IKISS system is based on an objected-oriented programming (OOP) architecture. To ensure the system is capable of federating image data with existing databases, such as caBIG [12] or the Allen Institute for Brain Science [15] (Figure 2) , the class, object, and composite structure were created in with a Unified Modeling Language (UML)based tool. JavaScript programming tools such as JQuery and the Scriptaculous library were used to create the necessary image interaction features. An interaction diagram, such as a sequence data, communication-link and

interaction overview diagram, was designed to retrieve image data using the IKISS programming interface.

3.3 INTERACTIVE IMAGE AND DATA PROCESSING FUNCTIONALITY As seen in Figures 3, 4 and 5 the IKISS system supports interactive image processing. Furthermore, IKISS uses a similar permission/access system to the Slim-Prim system [1, 2]. This allows image datasets to be shared when appropriate, and further allows the principal investigator/data manager to specify which data are visible and whether comments may be made. These image and data processing functions can be useful in preprocessing an image while still being online. The IKISS image processing functionality includes: 1. An image marker to identify and share ROI; 2.

Image view and pan functions to ease navigation of large images. This function prevents browser timeout from attempting to download large files;

3.

Side-by-side image comparison;

4.

A self-extracted expression data matrix that uses the K-Means Classifier automatic machine learning method to determine the expression-intensity distribution.

5.

Singular value decomposition computation;

6.

Correlation matrix estimation.

As seen in Figure 3, a rectangular box is used as a marker tool for regions of interest (ROI). The collaboration and discussion zone are divided into three parts: 1. "Add Comment" allows a user to post or reply to questions or comments; 2.

A chat (instant message) function can be used for real-time online collaboration;

3.

Highlight Tool to add emphasis on to any word or phase.

3.4. IKISS SECURITY AND INFRASTRUCTURE IKISS is a plug-in modular adaptor to our Slim-Prim database. As such, each Principal Investigator (PI) strictly controls access to their own data. The PI is administrator of their “data module” within Slim-Prim and is given full administrative privileges. From an access control panel on the Slim-Prim administrative site a PI may designate staff or collaborators as Admin, Superuser or generate their own lab-specific roles. Within this “Permisssions Module” a PI controls read/write/edit prividges down to the level of individual variables. IKISS utilizes this rolebased system to allow each PI to specify who has access to each image or group of images, and what actions they

JOURNAL OF COMPUTER SCIENCE AND ENGINEERING, VOLUME 2, ISSUE 2 AUGUST 2010 4

can perform. Thus a PI can control IKISS actions including ROI mapping, ROI comment, discussion highlighting, and real-time chat. In addition to access control, we have deployed OpenSSL encryption on the database to protect data covered under HIPAA as well as transparent data encryption (TDE) on the oracle server.

image search engines and image-mining approaches. Ultimately the development of the IKISS project is to assist investigators in basic science and clinical research studies with their long-term goals of working and sharing data with the wider research community. With novel technolgies such as IKISS, knowledge management and just-in-time information processing can support and promote healthcare decision making.

3.5. IKISS DEPLOYMENT AND UTILIZATION The IKISS system has been successfully deployed at UTHSC and currently interfaces with the Slim-Prim system [1, 2]. The API can be customized to interface with external systems to retrieve and import images for analysis and comparison, for example retrieval of brain images is possible from Allen Institute for Brain Science [15] (Figure 2). Furthermore we have built extensive additional functionality into the IKISS system such as side-by-side comparison, annotations features and “instant chat” functions for online real-time virtual collaboration. Images can be annotated in the IKISS system and keywords can be used to search the National Library of Medicine PubMed system [16] for abstracts featuring these keywords. IKISS is being expanded to include server mounted advanced statistical software packages such as SAS and MATLAB neural network toolbox for gene expression analysis.

4. DISCUSSION IKISS is a web-based repository for biomedical images. We aimed to create a user-friendly system where researchers can store, share and analyze images all in one location. The remit of the UT-BMIU is help break down the barriers separating research silos and to foster collaboration between basic and clinical scientists. We believe the IKISS system help create such an environment.The image management system used for the IKISS was originally created to organize basic static images. We have extended it to include many interactive and preprocessing image functionalities (see Section 3.1 and 3.3). IKISS makes it possible to view and annotate images online and then share these within a research group. Online discussion tools allow or real-time collaboration. We envision such tools being especially useful for fragmented research groups where laboratories may be spread between campuses or even across multiple time zones. IKISS is a modular expansion of the current Slim-Prim system and as such as we have no intention of reinventing advanced statistical software packages, such as SAS and image processing tools such as those in the MATLAB toolbox. On the contrary, we have developed a seamless interface to bridge with those software packages (see Section 3.5). We allow users to preprocess their biomedical images, and then isolate those extracted datasets for external examination. The system is expandable, and we envision great utility for biomedical research projects where different modalities are common within an individual project. Further studies in basic science and clinical research at UTHSC will be evaluated to guide the future expansion of IKISS. We plan on developing service-oriented biomedical

Fig. 3 Region of Interest(ROI)

JOURNAL OF COMPUTER SCIENCE AND ENGINEERING, VOLUME 2, ISSUE 2 AUGUST 2010 5

5. CONCLUSION Development of a system that can tackle vast amount of heterogeneous biomedical image modalities in one centralized location was a challenging, but rewarding task. The IKISS project has successfully demonstrated the first stage of managing, interacting, preprocessing, and sharing biomedical images between basic and clinical researchers in a userfriendly, versatile and secure web based system. The IKISS project is currently active in several basic and clinical research studies at UTHSC.

ACKNOWLEDGMENTS This work was supported by funds to the Biomedical Informatics Unit at the University of Tennessee Clinical Translational Science Institute (CTSI). The authors thank Dr. Michael P. McDonald for use of images.

REFERENCES [1]

Fig. 4. Image pan,zoom,rotation on a sagittal view

Fig. 5.Image interraction feature: discussion highlight on xrays image

T. Viangteeravat, I.M. Brooks, E.J. Smith, N. Furlotte, S. Vuthipadadon, R. Reynolds, and C.S. McDonald, “Slim-Prim: A Biomedical Informatics Database To Promote Translational Research,” Perspectives in Health Information Management, vol. 6, no. 6, May 2008. [2] T. Viangteeravat, I.M. Brooks, W.J. Ketcherside, R. Homayouni, N. Furlotte, S. Vuthipadadon, and C.S. McDonald, “Biomedical Informatics Unit (BMIU): Slim-Prim System Bridges The Gap Between Laboratory Discovery And Practice,” Clinical and Translational Science, vol. 2, no. 3, June 2008. [3] P.A. Harris. R. Taylor, R. Thielke, J. Payne, N. Gonzalez and J.G. Conde, “Research Electronic Data Capture (REDCap) - A MetadataDriven Methodology And Workflow Process For Providing Translational Research Informatics Support,” Journal of Biomedical Informatics, vol. 42, no. 2, April 2008. [4] P.J. Slomka, E. Elliott, and A.A. Driedger, “Java-Based PACS And Reporting System For Nuclear Medicine,” Proc of Medical Imaging 2000:PACS Design and Evaluation: Engineering and Clinical Issues, vol. 3870, February, 2000. doi:9.1016/11.376408 [5] PHP Group, “PHP Hypertext Preprocessor”. http://www.php.net/ (retrieved June 209) [6] P.J. Slomka, E. Elliott, and A.A. Driedger, “Java-Based Remote Viewing And Processing Of Nuclear Medicine Images: Towards ‘The Imaging Department Without Walls’,” Journal of Nuclear Medicine, vol. 41, no. 1, January 2000. [7] J.I. Kim, D. D. Feng, and T.W. Cai, “A Web Based Medical Image Data Processing And Management System,” ACM International Conference Proceeding Series, vol. 8, 2000. [8] T. Kalinski, R. Zwnitzer, T. Jonczyk-Weber, H. Hofmann, J. Bernarding, and A. Roessner, “Improvements In Education In Pathology: Virtual 3D Specimens,” Pathology, Research and Practice, vol. 205, no. 11, May 2008. [9] [R. Zwönitzer, T. Kalinski, H. Hofmann, A. Roessner, and J. Bernarding, “Digital Pathology: Dicom-Conform Draft, Testbed, And First Results,” Computer Methods and Programs in Biomedicine, vol. 77, no. 3, July 2007. [10] Oracle Corporation. “MySQL Database Engine,” http://www.mysql.com (Retrieved June 209). [11] D.B. Keator, J.S. Grethe, D. Marcus, B. Ozyurt, S. Gadde, S. Murphy, S. Pieper, D. Greve, R. Notestine, H.J. Bockholt, and P. Papadopoulos, “A National Human Neuroimaging Collaboratory Enabled by the Biomedical Informatics Research Network (BIRN),” IEEE Transac-

JOURNAL OF COMPUTER SCIENCE AND ENGINEERING, VOLUME 2, ISSUE 2 AUGUST 2010 6

[12] [13] [14] [15] [16]

tions on Information Technology in Biomedicine, Special Bio-Grid edition, vol. 11, no. 2, March 2007. National Cancer Institute. “Cancer Biomedical Informatics Grid caBIG,” https://cabig.nci.nih.gov/ (Retrieved June 209). Pixelpost, “Pixelpost,” http://www.pixelpost.org/ (Retrieved June 209). L. Zhang, L. Chen, F. Jing, K. Deng, and W.Y. Ma. “Enjoyphoto: A Vertical Image Search Engine for Enjoying High-Quality Photos,” Proc 13th annual ACM international conference on Multimedia, 2006. Allen Institute for Brain Science, “Allen Institute for Brain Science,” http://www.alleninstitute.org/ (Retrieved June 209). U.S. National Library of Medicine, National Institutes of Health “PubMed,” http://www.ncbi.nlm.nih.gov/pubmed/ (Retrieved June 209).

T. Viangteeravat, Ph.D. is currently an Assistant Professor with the Department of Preventive Medicine at the University of Tennessee Health Science Center. He is a member of the SPIE international society, DSP Elsevier, and Health Level Seven International (HL7.org). His research interests include interactive technology, software development, multimedia instruction, computer-based design, as well as sensor fusion, data compression, detection and estimation in one dimensional and multi-dimensional data problems. M.N. Anyanwu, Ph.D. is currently a Health Informatics Specialist with the BMIU of the UT-CTSI. He is a member of IEEE, ACM, ISA and the National Society of Black Engineers. He is an editorial board member of Journal of Global Research in Computer Science. His research interests include biomedical informatics, bioinformatics, data/text miming, software development and design interactive technology, database technology, business intelligence/decision support systems and multimedia interactive systems. V. Nagisetty, M.S. is currently a Research Data Analyst with the BMIU of the UT-CTSI. His research interests include software development, security management in health information protection, application of artificial intelligence for quality control in biomedical and clinical information. E. Kuscu, M.E. is currently a Research Data Analyst with the BMIU of the UT-CTSI. He holds a Bachelor of Science degree in Computer Engineering, Sakarya University, Turkey and a Masters degree in engineering with specialization in Robotics, Tennessee State University (2008). His research interest includes biomedical informatics, database management system, software development and design. I.M. Brooks, Ph.D. is the Program Manager of the BMIU of the UTCTSI. He received his Ph.D. in Biology from Penn State University in 2003. He is a member of SfN and AAAS. His research interests focus on use of public and non-public databases for improving health outcomes. C.S. McDonald, Ph.D. is Associate Director of the BMIU of the UTCTSI, Assistant Vice-Chancellor of the Office of Educational Technology & Institutional Research, and an Associate Professor in the Dept. of Preventive Medicine. Her research interests include software development, multimedia instruction, and computer-based design in biomedical research.