Multi-modal imaging can be integrated for a research enterprise using database and web technologies.

In this abstract we present a method to archive and manage multi-modal/multi-scale datasets within a large imaging research environment. This work stems from our own need for proper informatics in imaging science research. However, our work extends seamlessly to the growing need for data management, in light of growing algorithmic advances and multi-modal imaging datasets that are becoming pervasive in medical imaging science.[1] From the perspective of translational research, the need to manage both pre-clinical and clinical information becomes very important.[2,3,4]

We have developed a database system that is extensible to the various projects and datasets that are created within our research environment at the Vanderbilt University Institute of Imaging Science (VUIIS). The VUIIS contains over 30 image acquisition systems and nearly 100 researchers involved with a variety of research programs encompassing pre-clinical and clinical data. Working in this environment demands a system that is able to accommodate the variety of data collected, as well as diverse research designs and experiments. Another important consideration in developing our informatics toolset was the need to handle data analysis and processing information, along with derivative datasets born out of our image analysis.

While the goals of this work have been provoked by requirements within our research environment, the tools and techniques described in this abstract have broad applicability within general medical informatics.

We have developed our integrated PACS solution using MySQL as the database backend, and with two GUI platforms for end-user interaction. The database schema has been developed with extensibility at its core. The key elements of our database include master tables with records for individual projects, people working on each project, the datasets associated with each project, and the attributes for each dataset. These key elements are tied together in one-to-one and one-to-many relationships using linking tables and records. A feature of our schema is that other key elements can be integrated into our informational structure without affecting the existing design and data.

For end-user interaction we have chosen to develop two GUI platforms, one for web deployment and the other as an installable client application. For web interaction we have chosen to incorporate elements of the Google Web Toolkit (GWT), which provides a Java to Javascript compiler, along with a basic set of Java libraries and an API capable of generating GUI elements for web interfaces. The GWT also allows the incorporation of server-side interactions via scripts written in PHP, by passing data objects asynchronously using Javascript Object Notation (JSON). Thus, our complete web interface uses GWT for the GUI elements to manage user interactions, and to coordinate asynchronous data communication with server-side elements written in PHP.

As an extension to the web browser interface, we have developed a standalone application written in Python. Python presents many advantages, including a vast array of libraries tuned for specific functions, as well as cross-platform compatibility. By leveraging these benefits, this stand-alone interface allows us to deliver more computationally intensive interactions such as three-dimensional renderings or real-time image analysis from within the database interface. We employed the Qt GUI framework to provide the interactive elements within the application. Server-side interaction was accomplished using the same JSON interface and PHP scripts developed for the web application.

Currently there are 19 active projects on the vuPACS database system with over 80 different possible attributes and over 30 users. In this section, we will highlight 4 different projects that demonstrate the breadth of research utilizing our infrastructure.

Two projects which are relatively self-contained within the institute include one which aims to track therapeutic efficacy of molecular imaging agents in light of pre-clinical models and dosage distributions. Another aims to investigate parameters of tissue perfusion and cellular function. In the former, the VUIIS PACS infrastructure is used to coordinate and track datasets from CT and PET imaging techniques in corresponding animals over time. The dataset attributes track the animal ID, the model of disease, the dosage of therapy and the imaging data, over time. In the latter, PET images (highlighting cellular proliferation) in tumor bearing animals are tracked over time along with corresponding MR information of apparent diffusion coefficients (tissue cellularity). This project is unique in that derivative datasets (co-registered PET and MR data) are also tracked for each animal and time-point.

Another project that highlights the intra-mural collaboration capabilities of our developed infrastructure is one that utilizes in vivo MR data along with ex vivo MALDI imaging data. MALDI imaging is a relatively recent imaging technology, developed at Vanderbilt University at the Mass Spectrometry Research Center (MSRC). The purpose of this technology is to provide tissue proteomic information in situ, which we aim to correlate with in vivo characteristics of tissue structure and function from MR images. This project includes 5 different dataset types within the PACS system, including MR and MALDI native data, blockface imaging data (used for co-registration), derivative quantitative data from MR scans (T1 and T2), and derivative co-registered data. Each dataset type is managed by individuals spanning multiple positions and departments, including research assistants, graduate students, and post-doctoral fellows within the VUIIS and MSRC. This project demonstrates the versatility of the developed infrastructure in successful collaborative research involving many different dataset types, people, and departments.

The final project highlights the ability of our PACS system to provide service-oriented project tracking. In this project the principal investigator is an external collaborator (in the department of cardiac surgery) who utilizes both our pre-clinical imaging facilities and data analysis center to provide data acquisition and processing in a service model. The goal is to use MR to image cardiac aneurysm development, treatment, and therapeutic monitoring. The PACS infrastructure is used to house the MR images, as well as derivative information from co-registration and data analysis of vessel occlusion over time in corresponding animals.

Clinical projects have also been incorporated into our infrastructure. Many are related to human fMRI data, task paradigms, and their derivative results. These projects are forthcoming and their utiliziation of our developments is currently under investigation.

We have demonstrated a technique to coordinate and manage the variety of imaging science research within a large-scale, trans-institutional research enterprise. Our solution has been used effectively to track and manage datasets generated within our imaging facilities, while easily extending to other research centers within our institution. Our solution allows for efficient extension to a variety of imaging methods, parameters, as well as a multitude of derivative imaging metrics. The developed infrastructure scales well with various departments and allows for input from a variety of individuals working on each project. We believe that our solution is capable of providing critical data management and capable of growing as our institution grows. Our future work is aimed at further refinements to the backend processing infrastructure (including asynchrohous data packaging and processing) as well as enhacements to our GUI clients (to include advanced rendering and processing capabilities in real-time). We believe the developed technology has broad impact to other research centers interested in multi-modal, multi-scale imaging informatics for research purposes.

(1) Benaron DA. The Future of Cancer Imaging. Cancer and Metastasis Reviews. March 2002;21(1):45-78.

(2) Taylor MD, Mainprize TG, Rutka JT. Bioinformatics in neurosurgery. Neurosurgery. April 2003;52(4):723-30.

(3) Lowe HJ, Buchanan BG, Cooper GF, Vries JK. “Building a medical multimedia database system to integrate clinical information: an application of high-performance computing and communications technology.” Bulletin of the Medical Library Association. 1995;83(1):57–64.

(4) Wong ST, Huang HK. “Design methods and architectural issues of integrated medical image data base systems.” Comput Med Imaging Graph. July-August 1996;20(4):285-99.