Research Background
The fast development of computer and Internet technologies support the progresses of medical Information Technology (IT) greatly, which is very important for the construction of national healthcare information infrastructure. But now healthcare IT encounters big challenges to integrate the heterogeneous medical information systems (e.g. HIS, PACS and CIS) and application systems (e.g. medical image retrieval, 3D reconstruction and computer aided diagnosis system) distributed widely in different departments and hospitals. That restricts the medical research and commercial applications within one limited area. And the usage efficiency of related systems is kept at low level. Our research of MedImGrid (Medical Image Grid) aims to combine healthcare IT, SOC (Service Oriented Computing) and Grid technology to tackle the bottlenecks of healthcare infrastructure construction. The success of MedImGrid will help to decrease the cost of medical treatment and research, and bring us a new borderless digital medical-service time.

Current Research
Our group selects 3 typical applications as below within medical domain to tackle. Furthermore our some ongoing works to whole MedImGrid platform aim at adopting some virtualization means to realize one service-independent grid platform to improve the service management, discovery and composition within grid environment that mass web services are integrated in.

Medical Information Integration:
Now the medical information system integration within grid can be realized easily based on database access and integration middleware as CGSPP-DAI and OGSA-DAI (adopted by e-Diamond). But that means may threat the medical data security and can not combine with the current local information system access means. So our group now works hard to merge the EPR (Electronic Patient Record) exchange standard, ANSI-accredited HL7 (Health Level 7), in our grid middleware to enable the grid platform can access the medical information systems as virtual client. And related data mode mapping of heterogeneous information systems based on semantic also makes some progresses, which will be adopted within our platform in the future.

EPR Retrieval Result in Different Hospital Node

Medical Image Online Workstation

HL7-RIM Ontology Tree

3D Reconstruction Based on Tomographies
Medical images, which present the visual healthy or diseased features, are very important to assist doctors’ diagnosis. But under current conditions, the tomography images captured by different devices (such as CT, PET and MRI) can not be read easily. Only experienced clinicians can find out abnormal images from mass cases. Some rare details of medical images that go beyond of the clinicians’ cognizing space will be ignored regrettably. Medical image processing technology (such as 3D reconstruction, images merging and surgical operation simulation) strengthens the feature details of images or transforms them to the forms that people can observe conveniently. But related tools are bound with high-cost hardware and software, and are special for fixed diseases and image types. That restricts them from being used widely. Our group is working hard to improve the parallel computing ability of medical image processing algorithms to exert the superiorities of grid computing. 3D reconstruction can create 3D surface models based on tomography images. The algorithm makes it possible that the clinicians can observe medical images with an approximate view angle resembled with human knowledge.

3D-Reconstruction Interface

3D-reconstruction Effect

Medical Image Retrieval
Typical medical images, which present the visual healthy or diseased features with concrete samples, are very important to assist the training and diagnosis of doctors. Many physicians construct their medical image libraries independently, which store representative case samples collected in a long time with detailed disease background and evolvement information. Some CBIR products have appeared in medical device market with high hardware cost and high price that restricts them from being popularized widely. Furthermore because the limits of geographical locations and clinical experiences, the image libraries of different specialists can not cover all possible disease features with limited samples. So our group aims at merging grid technology to realize the integration of divided libraries, and make the images and their medical background information shared across the border of departments and hospitals.

Segmentation Process

Doubted Areas

CSBIR Retrieval Result