HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Pan, T. C. Articles by Siegel, E. L. Search for Related Content PubMed PubMed Citation Articles by Pan, T. C. Articles by Siegel, E. L. Related Collections Informatics

Informatics in Radiology

GridCAD: Grid-based Computer-aided Detection System¹

¹ From the Department of Biomedical Informatics, Ohio State University, 3190 Graves Hall, 333 W 10th Ave, Columbus, OH 43210 (T.C.P., M.N.G., S.A.L., S.W.O., S.L.H., A.S., B.G.R., D.W.E., T.M.K., J.H.S.); VA Maryland Health Care System, Baltimore, Md (K.M.S., E.L.S.); and University of Maryland School of Medicine, Baltimore, Md (E.L.S.). Presented as an infoRAD exhibit at the 2005 RSNA Annual Meeting. Received August 17, 2006; revision requested September 22 and received November 8; accepted December 20. Supported in part by the National Cancer Institute, the National Science Foundation (CNS-0509326, CNS-0403342, ANI-0330612), the National Institutes of Health (NIBIB BISTI P20EB000591), and the Ohio Board of Regents (BRTTC BRTT02– 0003, ODOD-AGMT-TECH-04 – 049). M.N.G. is a stockholder in iCAD. K.M.S. is a speaker for TeraRecon, San Mateo, Calif; cofounder of iVirtuoso, Baltimore, Md; and a member of the advisory board of GE Healthcare IT, Barrington, Ill. E.L.S. received research funding from GE Healthcare. All other authors have no financial relationships to disclose. Address correspondence to T.C.P. (e-mail: tpan{at}bmi.osu.edu).

	Abstract

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
Grid computing—the use of a distributed network of electronic resources to cooperatively perform subsets of computationally intensive tasks—may help improve the speed and accuracy of radiologic image interpretation by enabling collaborative computer-based and human readings. GridCAD, a software application developed by using the National Cancer Institute Cancer Biomedical Informatics Grid architecture, implements the fundamental elements of grid computing and demonstrates the potential benefits of grid technology for medical imaging. It allows users to query local and remote image databases, view images, and simultaneously run multiple computer-assisted detection (CAD) algorithms on the images selected. The prototype CAD systems that are incorporated in the software application are designed for the detection of lung nodules on thoracic computed tomographic images. GridCAD displays the original full-resolution images with an overlay of nodule candidates detected by the CAD algorithms, by human observers, or by a combination of both types of readers. With an underlying framework that is computer platform independent and scalable to the task, the software application can support local and long-distance collaboration in both research and clinical practice through the efficient, secure, and reliable sharing of resources for image data mining, analysis, and archiving.

© RSNA, 2007

	Introduction

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
The Internet originally was developed to meet requirements for communication among multiple federally funded computing centers. Requirements for resource sharing and communication among these groups underscored the necessity to standardize the protocols needed to exchange information. The continuing development of these protocols evolved into what we now think of as the Internet, which has transformed the way in which information is shared by hundreds of millions of users around the world. Grid computing is an emerging technology that offers an infrastructure for sharing not only information but also geographically dispersed computational and storage resources. The grid has been defined in the computer literature as an environment that facilitates "flexible, secure, coordinated resource sharing among dynamic collections of individuals, institutions, and resources" (1). It can be thought of as an extension of the framework of the Internet to create an even more generic and powerful resource-sharing environment.

Grid computing grew out of the work of many computer scientists in the 1980s and 1990s. Today, the most widely accepted grid software system is the Globus toolkit, which was developed by Ian Foster, Carl Kesselman, and Steve Tuecke. The Globus toolkit is an open-source framework for computer processing and storage management, security, data movement, and monitoring. Unlike traditional computer clusters or distributed computing, grid computing provides support for computation across multiple disparate administrative domains. Grid computing makes it possible to unite resources from different computer platforms, with different architectures, using different computer languages, and in multiple locations, over a single network by using open standards. This enables what has been referred to as virtualization of computing resources. It provides efficient and secure ways to share resources, including data, software applications, computational capabilities, and storage capacity, by using open protocols and standardized service interfaces. It creates unprecedented possibilities for new forms of collaborative investigation as well as more powerful clinical and research applications. The Globus toolkit has been used to support a wide variety of applications in the physical sciences, engineering, and biomedicine.

Perhaps the best-known application of grid computing is the project SETI@home (http://setiweb.ssl.berkeley.edu), in which hundreds of thousands of personal computers with many different software platforms (including Microsoft Windows, Linux, and Macintosh operating systems) are connected via the Internet in a cooperative and coordinated search for radio signals with nonnatural patterns that might indicate extraterrestrial intelligence. The application uses data collected by the Arecibo radio telescope in Puerto Rico and employs the Berkeley Open Infrastructure for Network Computing grid toolkit. The application coordinates the processing of the complex radio telescope signals by dividing them into multiple small segments that can be analyzed individually on personal computers using the SETI@home screen saver programs.

Grid computing is particularly well suited to complex and computationally demanding applications in medical imaging, such as computer-aided detection (CAD). Especially intriguing is the possibility of using CAD software programs from different vendors in a cooperative manner to enhance the performance and accuracy of lung nodule detection in a single image data set. The nodule candidates selected in a composite automated reading then could be combined or compared with those identified by one or more human observers for research purposes or for clinical image interpretation.

Lung cancer is currently the most common cause of cancer deaths among both men and women. According to a recent report from the American Cancer Society (2), more people in the United States died of this disease in 2005 than of breast, prostate, and colon cancers combined. A number of authors have suggested that a substantial percentage of clinically significant lung lesions missed in routine clinical interpretation of thoracic computed tomographic (CT) studies might be detected with the use of CAD systems (3–7). This technology is increasingly being applied to other imaging modalities and disease states as well. For example, CAD systems are used for the detection of masses and microcalcifications at mammography, lung nodules at chest radiography, and polyps at CT colonography (so-called virtual colonoscopy) (8,9). CAD algorithms have improved in accuracy and ease of use, while medical image data sets rapidly have increased in size, making such algorithms increasingly valuable in clinical practice. Changes in reimbursement also are spurring the rapid incorporation of CAD algorithms into routine practice, as is a growing body of literature supporting the ability of CAD systems to increase diagnostic accuracy (particularly sensitivity) when used in combination with human readers. However, the additional time required for use of CAD systems suggests a need for the development of modifications in workflow and in how CAD is used to streamline the interpretation process.

CAD can be thought of as a computer vision system that uses advanced pattern recognition and image analysis techniques to automatically detect medical abnormalities. Current commercially available and experimental CAD systems operate on local data sources. In most practices, a CAD system from a single vendor is used at a specific location. In this article, we describe gridCAD, a software system that integrates into a grid framework different CAD programs from multiple vendors, thereby creating an infrastructure that allows invocation of multiple CAD algorithms in parallel on one or more image data sets. The innovative use of grid computation in gridCAD offers the potential to greatly increase the accuracy and speed of image analysis by sharing data as well as computational resources. This approach also enables the creation of a consensus among multiple CAD systems and the combination of the CAD system–based interpretation with interpretations from one or more radiologists in one or more locations.

	Anatomy of a Health Care Grid

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
A generic grid (Fig 1) may be composed of one or more user interfaces or applications and several types of grid services, such as data services, analysis services, computing services, and middleware support services. Grid-based applications connect and interact with these different components to accomplish various analysis tasks.

View larger version (37K): [in this window] [in a new window] [Download PPT slide]   Figure 1.  Schematic shows the architecture of a hypothetical health care grid.

The infrastructure provided by the health care grid data service allows a researcher, radiologist, or clinician to easily create, manage, and advertise the availability of new data; to search within existing data sources; to query and retrieve interesting data subsets; and, finally, to integrate the retrieved data according to specified research or clinical requirements. The retrieved data conform to well-defined published standard schemas, and their conformity obviates user knowledge of the particular data formats used by different researchers and clinicians.

Analysis Services
Once data (eg, CT images of the thorax) have been integrated into a virtual data repository, they can be processed to extract meaningful information. The algorithms, tools, and applications that perform image analysis and data mining operations are advertised as grid analysis services that are shared across the Internet. Existing applications can be incorporated into the grid as analysis services, or new tools may be developed, to leverage the distributed processing capability of the grid.

Computing Services
As data sets increase in size and resolution and as algorithms increase in sophistication, computational requirements may exceed local storage and computational capacities. In such cases, "compute farms," which consist of dedicated systems for computation, can enable large-scale analysis. Compute farms can advertise their computational capabilities on a grid. Users can direct their data and applications to these farms and can access the farms’ resources according to a preassigned level of authorization. Users and institutions can use grid computing to pool data, applications, and computing resources, thus creating a better and more collaborative research environment. Alternatively, users can offer excess computing capacity back to the grid, in a process analogous to an individual selling excess solar electricity to the power company on a supply grid.

Middleware Support Services
The distribution of resources within a grid necessitates means of locating services within that grid environment. Figure 1 shows the service registry, semantic registry, protocol registry, and security infrastructure, which are key components of the generic grid middleware infrastructure. The service registry provides a way for data and analysis services to advertise their existence and their capabilities. A user then can locate the appropriate services on the basis of capabilities, hardware and software requirements, input and output data formats, and parameters. Once the service location and invocation parameters are identified, the service can be invoked by using a properly formatted request.

Data and analysis service requests and responses must follow standard published protocols to ensure interoperability between the client and the service as well as among services. The protocols are published in globally and publicly accessible protocol registries. Communication protocols are typically specified by extensible markup language (XML) schemas, and requests and responses are transmitted as XML instance documents. Some grids (eg, the Cancer Biomedical Informatics Grid [caBIG]) contain semantic registries that are designed to manage vocabulary and data elements used by grid applications and services. These vocabularies and common data elements generally include controlled vocabularies and data models produced by various medical application communities and enable much more complex data mining and analysis than would be possible otherwise.

Data and algorithm sharing in a health care environment requires careful planning for security infrastructure. It is the responsibility of the security service to maintain proper access to the data through user and analysis service authentication and authorization. Some data may include patient information and thus must conform to requirements of the Health Insurance Portability and Accountability Act (HIPAA) and the Joint Commission on Accreditation of Healthcare Organizations, as well as specific requirements of individual states and institutional review board–approved research protocols. In a research setting, the security service also may provide the deletion of patient-identifying information from clinical data. The security service can be used to restrict access to proprietary analysis services to specified personnel.

Health Care Grid Application User Interfaces
The components of the grid are linked via an application-specific user interface (Fig 2). The user interface defines workflow and, consequently, the flow of data among different components in the grid environment. The user interface also provides mechanisms for retrieving and reviewing data and analysis results.

View larger version (34K): [in this window] [in a new window] [Download PPT slide]   Figure 2.  Schematic shows the gridCAD framework and the flow of information among the components.

	Data Sharing and Aggregation

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
Grid infrastructure provides a unified way to access remote data. In the health care setting, this can simplify data sharing among institutions and facilitate remote review of cases by multiple users. In addition, a clinical case or research study may involve subjects from multiple sites. For example, a researcher may want to analyze the prevalence of a given finding (eg, ground-glass opacities in patients less than 50 years old) for one or more specified geographic locations. Analyses can be directed at multi-institutional data sets without explicitly copying and aggregating the images. Each site also can stage analysis services that operate on local data and can transmit to the requesting researcher only aggregated results that maintain patient anonymity and confidentiality, thereby simplifying compliance with HIPAA and institutional review board requirements.

	Cancer Biomedical Informatics Grid

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
One of the most prominent and promising recent efforts in health care grid computing was the launch of caBIG, a community initiative sponsored by the National Cancer Institute (NCI) Center for Bioinformatics to create an informatics infrastructure among clinical cancer centers that would facilitate research through the sharing of data, software, and expertise. The NCI-funded caBIG program aims to improve cancer research and patient care through effective collaboration across institutions and disciplines in an open, standards-based environment. This includes common standards and data sets, grid infrastructure (caGrid) that extends grid middleware to support semantic and syntactic interoperability, and interoperable grid application for the participants. In April 2005, the In Vivo Imaging Work-space, a community within the caBIG program, was created to foster the development of projects that are designed to advance medical imaging informatics, especially in cancer care.

GridCAD, the software application described in this article and demonstrated at the 2005 RSNA Scientific Assembly and Annual Meeting, was constructed by using caGrid, which was developed as part of the caBIG initiative (Fig 2). CaGrid is a middleware infrastructure that provides a communications layer for applications to interact across different hardware and network environments. It is also a toolkit that supports the development of grid-enabled, caBIG-compliant applications. The caGrid toolkit leverages the Globus toolkit, NCI Cancer Data Standards Repository, Global Model Exchange (an XML schema management system produced by the Mobius project), and Open Grid Services Architecture–Data Access Integration. The Globus toolkit provides the following set of core components for the development of grid applications (10): security, including authentication, authorization, and credential management; data management, including data transfer and associated optimizations; execution management and resource allocation; information services for monitoring and discovery; and a common runtime library for application development support.

	GridCAD Implementation

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
To demonstrate the benefits of grid computing for CAD, we implemented gridCAD for the detection of lung cancer. GridCAD differs from other grid services (see "Summary and Future Prospects") in its use of community standard toolkits, caGrid, and community-accepted data communications standards, as well as in its goal of providing a common and open framework for rapidly creating and deploying grid-based image analysis applications.

As discussed previously, a grid computing environment has several components, including data services, analysis services, middleware support services, computing infrastructure, and user interfaces. Each of these is implemented and utilized in gridCAD for the lung cancer application. The objectives of the gridCAD framework are achieved by exposing a CAD algorithm as a grid-aware service and by facilitating the easy and secure exchange of images and CAD results. One approach for exposing an application or a data source as a grid service is to wrap it inside a layer that facilitates interaction and communication with other grid components while leaving the original unmodified. The wrappers that we used come from the caGrid toolkit. In gridCAD (Fig 2), the following grid components are implemented: CAD analysis services, which invoke CAD systems and manage the flow of data; image data services, which provide interfaces to the data repositories (eg, a picture archiving and communication system [PACS] server); middleware support services, which provide operational support such as storage and communication schemas, data security, application invocation, and CAD result storage (in repositories such as the Mako XML database, a product of the open-source Mobius project); and user interfaces, which allow query, original image preview, and CAD system marking review.

	Reference Implementation in Lung Cancer Detection

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 
In our implementation of gridCAD for lung cancer detection, we used CT image data sets from the Reference Image Database to Evaluate Response (RIDER), an archive of time-series images of lung cancer. We also included a thoracic CT database of 100 patients at the Baltimore Veterans Affairs Medical Center. The RIDER data sets, which contain image data that were acquired with various imaging modalities and protocols at various institutions and from which patient-identifying information has been deleted, are available for free public use from the NCI National Cancer Imaging Archive Web site (http://ncia.nci.nih.gov). Information that might allow patient identification also was removed from the image data sets that we received from the Baltimore Veterans Affairs Medical Center, and we received an institutional review board exemption for our use of these data. The images from the Baltimore Veterans Affairs Medical Center were acquired and reconstructed at a section thickness of 0.75 mm. Image data were stored on five different Digital Imaging and Communications in Medicine (DICOM)-compliant PACS servers that were exposed to the grid as image data services that are geographically distributed across the United States. The locations of the gridCAD image data services included the infoRAD booth at the 2005 RSNA annual meeting, the Collaboratory for Advanced Computing and Simulations at the University of Southern California (Los Angeles, Calif), and the Multiscale Computing Laboratory at Ohio State University (Columbus, Ohio). The DICOM-compliant server software that was used to host and manage the images was the open-source PACS server software in the PixelMed Java toolkit library (http://www.dclunie.com/pixelmed/software). More than 800 image series were used for the gridCAD implementation.

CAD analysis services were implemented as wrappers for lung nodule CAD algorithms from Siemens Medical Solutions (Malvern, Pa) and iCAD (Nashua, NH). Each of these algorithms was treated as a black box with a well-defined command-line interface. A separate analysis service wrapper was developed for each CAD algorithm because the algorithms were not developed for a grid computing environment and have different interfaces for invocation and different format requirements for data input and output.

The outputs of the two CAD algorithms were presented to the user as overlays on the CT images. Each nodule candidate was marked with a circle or a square of different colors. The user could scroll through the images as well as change the intensity window and level during the review. Figure 3 shows the outputs of the two CAD algorithms. The demonstration of gridCAD at the 2005 RSNA annual meeting showed how a radiologist could incorporate this system into his or her routine workflow by performing image interpretation and then using a combination or consensus of markings from multiple CAD algorithms. The demonstration also illustrated how interpretations from other radiologists could be combined with the CAD system markings.

View larger version (165K): [in this window] [in a new window] [Download PPT slide]   Figure 3.  CT image with overlay shows different abnormalities identified by two CAD algorithms from different vendors. Findings marked by each algorithm are identifiable according to the geometric shape (circles or squares) and color (red or blue; here, shown in black) of the mark. In the gridCAD framework, the CAD systems are connected to the grid infrastructure as services, and the user can invoke more than one at a time.

	Summary and Future Prospects

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

The number of radiology-related applications of grid computing is increasing. Such applications include MammoGrid, a pan-European database that allows access to mammograms by using a grid-based software (11). One objective of the MammoGrid project is to extract tissue-level information (eg, the number and location of micro-calcifications) for use in clinical studies. In another European effort, a CAD system for mammographic analysis (Computer Assisted Library for Mammography) has been integrated with a grid-based mammographic reading environment for use in the detection of masses and microcalcifications (12). Grid-based CAD applications for the detection of breast cancer and Alzheimer disease also have been under development by the Medical Application on a Grid Infrastructure Connection–5 project group (http://www.magic5.unile.it).

One of the most intriguing practical applications of the grid in health care is its use as a mechanism to achieve unified access to multiple analysis services. In addition to cancer detection, tasks such as tumor volume measurement, therapy efficacy assessment, and parameter extraction from dynamic contrast material–enhanced magnetic resonance (MR) imaging data could be performed at the same time by using the most suitable algorithms.

The development and evaluation of effective algorithms require access to a large number of cases from different geographic locations so that variations in the population are adequately modeled. A traditional approach would involve the collection and transfer of image data to a central location. With the use of grid computing, widely distributed data are easily shared and accessible for development and evaluation. Once developed, image processing and analysis algorithms can be validated by using a large number of cases from several institutions before regulatory approval is sought. A grid system thus may facilitate a faster and hence more affordable regulatory approval process.

One of the primary goals of grid computing is to remove the dependence of an application on the underlying hardware and to deliver the application as a utility. As the complexity of a clinical or research imaging study increases, more computing resources can be recruited easily. The gridCAD system can provide this capability by means of CAD compute farms where dedicated clusters of high-performance computers run various CAD algorithms. A user or an institution can leverage these grid computing resources without having to support and maintain the computer equipment and can obtain resources that are scalable to the complexity of the task.

Moreover, grid support for algorithm and application workflows allows algorithm components to be assembled into data processing pipelines. A typical CAD program consists of several independent software modules and is built for a single computer. For example, a lung nodule detection program may consist of modules for the detection of lung contours, the segmentation of nodule candidates, and the reduction of false-positive nodule candidates. Different modules have different computational and memory requirements. The ability to rewrite these algorithms to support the distribution of computational modules and data storage across many computing resources has significant potential for improving the speed and performance of CAD and other complex algorithms.

The service-based architecture of gridCAD allows easy integration of additional algorithms and data sources by using caBIG tools and standards. This ease of integration facilitates the development and deployment of other disease-specific applications, such as mammographic CAD. Unified communication protocols allow additional data sources and analysis services to be added easily and in an ad hoc fashion.

In the current implementation of gridCAD, much of the focus has been on creating an operational infrastructure for running CAD algorithms on a grid. For clinical and research deployment, additional security features must be incorporated, including authentication and authorization of users and grid services, masking or deletion of confidential or patient-identifying information from clinical data, and secure transmission of information. Our primary implementation focus, consequently, is to address these features. Security in gridCAD will leverage an ongoing effort in the caBIG architecture workspace, specifically caGrid’s Grid User Management Service and Common Attribute Management Service. These services provide security and authentication capabilities, leveraging the Globus toolkit security components for management of users, their credentials, and usage permissions.

As the diagnostic imaging data obtained with CT and MR imaging increase in spatial and temporal resolution as well as in overall complexity, the amount of data stored per patient also increases dramatically. Latency in the transfer of data across the grid for remote image review increases with data size, thus adversely affecting the user’s ability to dynamically interact with databases. An efficient method of transferring large amounts of image data is important to the success of grid-based systems such as gridCAD. Efforts are already under way in the biomedical field to support large data transfers by using approaches such as data streaming, data compression with the Joint Photographic Experts Group 2000 Interactive Protocol, multiresolution data compression, and region-of-interest–based data transfer (the transfer of portions of images). However, some of these technologies have not yet been incorporated as use cases for caGrid’s communication protocols. High-performance data transfer remains an area of future work for gridCAD as well as caGrid.

Although high-performance data transfer may improve the transfer latency for an individual data set, large numbers of image data sets still present challenges for efficient data movement. Clinical trials that involve multiple sites, each with large numbers of image data sets, may further compound the problem. The resulting data transfer time on the grid can become a bottleneck for CAD system performance.

Moving algorithms to data repositories instead of moving data to analysis services reduces the amount of data transfer and may enhance overall system performance in certain scenarios (Fig 4). This requires a significant amount of middleware support for runtime transmission of algorithms and installation at remote grid services. We plan to implement this capability in future gridCAD versions, with the support of the caGrid toolkit and CAD system vendors.

View larger version (38K): [in this window] [in a new window] [Download PPT slide]   Figure 4.  Schematic shows the movement of algorithms to data repositories.

Grid computing has tremendous potential to create health care benefits; the medical imaging community has only begun to explore the possibilities. Promising grid computing applications include teleradiology services, distributed and remote image processing and analysis, quality assurance and research, and clinical data mining.

	TAKE-HOME POINTS

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 

Grid computing allows efficient and secure sharing of data, software applications, computational resources, and storage capacity by using open protocols and standardized service interfaces.

A generic grid may be composed of several services (eg, data, analysis, computation, and middleware support) and one or more user interfaces or applications that connect and interact with the grid components.

GridCAD is a software application that makes innovative use of grid computing to increase the speed and accuracy of radiologic image interpretation through the sharing of data and analysis resources.

GridCAD may be used to obtain a consensus interpretation by multiple CAD systems and human readers in one or more geographic locations.

GridCAD is built on the NCI Cancer Biomedical Informatics Grid (caBIG) architecture and is semantically and syntactically interoperable with services and applications in caBIG.

	Acknowledgments

 
The authors thank Dennis O’Dell, Marcos Salganicoff, Toshiro Kubota, and Rajesh Amara of Siemens Medical Solutions and Euvondia Friedmann, Tom Fister, Maha Sallam, and Tim Carter of iCAD for their assistance. The authors also thank the Collaboratory for Advanced Computing and Simulations at the University of Southern California for contributing storage resources.

	Footnotes

 

Abbreviations: CAD = computer-aided detection, HIPAA = Health Insurance Portability and Accountability Act, NCI = National Cancer Institute, PACS = picture archiving and communication system, RIDER = Reference Image Database to Evaluate Response, XML = extensible markup language

	References

Top Abstract Introduction Anatomy of a Health... Data Sharing and Aggregation Cancer Biomedical Informatics... GridCAD Implementation Reference Implementation in Lung... Summary and Future Prospects TAKE-HOME POINTS References

 

1. Foster I, Kesselman C, Tuecke S. The anatomy of the grid: enabling scalable virtual organizations. Int J Supercomput Appl 2001;15:200–222.
2. American Cancer Society. Cancer facts and figures 2005. American Cancer Society Web site. http://www.cancer.org/docroot/STT/content/STT_1x_Cancer_Facts__Figures_2005.asp. Published 2005. Accessed July 1, 2006.
3. Peldschus K, Herzog P, Wood SA, Cheema JI, Costello P, Schoepf UJ. Computer-aided diagnosis as a second reader: spectrum of findings in CT studies of the chest interpreted as normal. Chest 2005;128:1517–1523.[CrossRef][Medline]
4. Brown MS, Goldin JG, Rogers S, et al. Computer-aided lung nodule detection in CT: results of large-scale observer test. Acad Radiol 2005;12: 681–686.[CrossRef][Medline]
5. Farag AA, El-Baz A, Gimelfarb G, El-Ghar MA, Eldiasty T. Quantitative nodule detection in low dose chest CT scans: new template modeling and evaluation for CAD system design. Med Image Comput Assist Interv 2005;8(pt 1):720–728.
6. Gurcan MN, Sahiner B, Petrick N, et al. Lung nodule detection on thoracic computed tomography images: preliminary evaluation of a computer-aided diagnosis system. Med Phys 2002;29:2552–2558.[CrossRef][Medline]
7. Giger ML, Doi K, MacMahon H, Metz CE, Yin FF. Pulmonary nodules: computer-aided detection in digital chest images. RadioGraphics 1990; 10:41–51.[Abstract]
8. Summers RM, Jerebko AK, Franaszek M, Malley JD, Johnson CD. Colonic polyps: complementary role of computer-aided detection in CT colonography. Radiology 2002;225:391–399.[Abstract/Free Full Text]
9. Chan HP, Doi K, Vyborny CJ, et al. Improvement in radiologists’ detection of clustered microcalcifications on mammograms: the potential of computer-aided diagnosis. Invest Radiol 1990;25: 1102–1110.[CrossRef][Medline]
10. About the Globus toolkit. Globus Toolkit Web site. http://globus.org/toolkit/about.html. Accessed June 10, 2006.
11. Amendolia SR, Brady M, McClatchey R, Mulet-Parada M, Odeh M, Solomonides T. Mammo-Grid: large-scale distributed mammogram analysis. Stud Health Technol Inform 2003;95:194–199.[Medline]
12. Bottigli U, Cerello P, Delogu P, et al. A computer aided detection system for mammographic images implemented on a GRID infrastructure. Proceedings of the 13th IEEE-NPSS Real Time Conference 2003, Montreal, Canada, May 18–23, 2003.

This article has been cited by other articles:

				D. S. Mendelson, P. R. G. Bak, E. Menschik, and E. Siegel Informatics in Radiology: Image Exchange: IHE and the Evolution of Image Sharing RadioGraphics, September 4, 2008; (2008) 287085174. [Abstract] [Full Text]

				S. Langella, S. Hastings, S. Oster, T. Pan, A. Sharma, J. Permar, D. Ervin, B. B. Cambazoglu, T. Kurc, and J. Saltz Sharing Data and Analytical Resources Securely in a Biomedical Research Grid Environment J. Am. Med. Inform. Assoc., May 1, 2008; 15(3): 363 - 373. [Abstract] [Full Text] [PDF]

				S. Oster, S. Langella, S. Hastings, D. Ervin, R. Madduri, J. Phillips, T. Kurc, F. Siebenlist, P. Covitz, K. Shanbhag, et al. caGrid 1.0: An Enterprise Grid Infrastructure for Biomedical Research J. Am. Med. Inform. Assoc., March 1, 2008; 15(2): 138 - 149. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Pan, T. C. Articles by Siegel, E. L. Search for Related Content PubMed PubMed Citation Articles by Pan, T. C. Articles by Siegel, E. L. Related Collections Informatics