Menu Close

An Introduction to the Medical Grid Project

by Epifanio Bagarinao

Medical imaging technologies such as magnetic resonance imaging (MRI), computed tomography (CT), among others, have progressed considerably over the past several years. Obtaining sophisticated high-resolution images of the human body, non-invasive investigation of the different functions of the human brain, and excellent detection of diseases are but a few applications that are made possible by the development of these technologies. However, an explosion in the number of digital medical images also accompanies these advances. This presents new challenges in the management of these datasets in terms of the required storage and computational resources.

To answer the computational and storage demands of modern medicine, the Medical Grid Project (MGP) was initiated in 2004 with a funding from Japan’s Telecommunication Advancement Office (TAO) and the National Institute of Advanced Industrial Science and Technology (AIST). It is an international collaboration involving researchers from Japan, the Philippines, and recently Taiwan. Its aim is to study and demonstrate the idea of using grid technology for medical applications. For this, the MGP testbed was formed. The testbed provides three analysis servers (one PC cluster from Ateneo de Manila University (ADMU) in Quezon City, Philippines, one PC cluster from the National Center for Geriatrics and Gerontology (NCGG) in Nagoya, Japan, and one supercluster from the National Institute of Advanced Industrial Science and Technology (AIST) in Tsukuba, Japan) and two data servers (one in AIST and another one in NCGG). More details about the project can be found at the project’s website located at http://www.medgrid.org/. As an initial task, the project aims to use emerging grid technologies for the remote analysis of functional MRI (fMRI) datasets.

Functional MRI is a non-invasive tool used to investigate and study the functions of the human brain. In a typical fMRI experiment, the subject’s brain is continuously scan while the subject lying inside the MRI scanner performs some tasks. The resulting series of functional MR images is then collected and analyzed. Voxels (volume element) with intensity variation that correlates with the task design are considered active during the task. The analysis of fMRI datasets is a computationally intensive task that includes image preprocessing and statistical analysis. Image preprocessing usually includes realignment to correct for subject’s head motion during the scan, spatial normalization to align different brains to ideal or standardized brain for averaging or comparison, and spatial smoothing to increase the signal-to-noise ratio in the acquired functional images. Statistical analysis is employed to extract voxels that respond to experimental design, usually using the general linear model (GLM). The final result is what is called an activation map showing regions in the brain that are actively involved in the task performance.

To facilitate the analysis of fMRI datasets in a grid environment, the MGP testbed provides a software package called BAXGrid (brain activation explorer on grid). Written in C, the package is composed of a client component and a server component. The client component provides a graphical user interface (GUI), where users can interact, and runs in the user’s local workstation. The GUI allows users to specify, among others: 1) the scanning parameters such as the number of slices and the number of volumes to process, 2) the preprocessing operations that will be included in the analysis, and 3) the remote computational resources that will be used. It also provides an image viewer for the presentation of the resulting activation map. Through this GUI, the remote analysis is made transparent to the user, as if all computations are done using the local workstation. On the other hand, BAXGrid’s server component runs on the testbed’s remote analysis servers and does the actual processing of fMRI datasets. During the analysis, it receives data from the client component, processes the data, and sends the results back after processing.

The MGP testbed also provides BAXSQL, a software package designed for the management of fMRI datasets in a grid environment. It is entirely written in C, currently runs on Linux, and provides its own graphical user interface. For fMRI data management, it can be used to register new datasets or studies in the testbed’s data servers, view details of existing datasets, edit data information, browse available data, among others. Moreover, basic functional MR image pre-processing operations, such as realignment, smoothing, and normalization, and standard statistical analysis using GLM are incorporated into the application. Basically, the application calls the server component of the BAXGrid package to do the actual data processing. It also employs globus toolkit’s security infrastructure, allowing users access to remote resources without having to authenticate separately to each resources, thus making remote processing transparent to the users.

With the use of grid technology, MGP opens new possibilities to meet short-term computational demands and very large storage requirements. This is particularly useful for medical imaging facilities where computing and storage resources may not be economically cost effective to own and cannot be conveniently hosted and maintained locally. These facilities can simply connect to the MGP testbed when needed, thus enhancing their capabilities without the additional cost associated with resource acquisition and maintenance. Finally, MGP also enables researchers from different countries share their resources and collaborate with each other, thus promoting international research cooperation and hopefully producing more research output.

_____

For more information, please contact:

Epifanio Bagarinao
Grid Technology Research Center
National Institute of Advanced Industrial Science and Technology
Tsukuba City, Ibaraki, Japan
email: baggy @ bahaykuboresearch . net

 

 

Advertisements

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *