The Digital Imaging and Communications in Medicine (DICOM) standard was created to promote the distribution and viewing of medical images, such as computed tomography scans, ultrasound images, and magnetic resonance imaging files. Recently, the demand for rapid and mobile access to this medical information has risen dramatically. However, the rate of acquisition and display of high-resolution medical images is limited by the network transmission speed or the resources of handheld terminal devices. Some users may need a low-resolution image to view on a terminal device with a limited display screen, while others may be interested in viewing a specific region of the image in higher quality.
One approach is to store multiple versions of the same image on the server, but this method requires additional storage. Another is to have the server transcode the image before sending it to the client, but this wastes server computational resources. To address these problems, we present an approach based on the JPEG2000 Interactive Protocol (JPIP) to browse high-resolution medical images in a more efficient way.
The JPEG2000 image-compression standard offers many features that support interactive access to large images,1,2 including high-efficiency compression, resolution scalability, quality scalability, and spatial random access. JPIP is the interactive protocol standard for viewing JPEG2000 images in a client-sever system.3 It uses the scalable features of the JPEG2000 code-stream, allowing the client to instantly fetch the region of interest (ROI) without directly accessing the compressed target file.3,4
Based on JPIP, we designed and implemented an interactive-image communication system with a client-server architecture (Figure 1). The JPIP client communicates with the JPIP server through an intranet or Internet connection, and personal digital assistants (PDAs) can access the server over a wireless network. The client sends Hypertext Transfer Protocol-GET requests to the server. The server retrieves the requested images from the picture archiving and communication system (PACS) server using a WADO (Web Access DICOM Persistent Object) protocol, stores them in a local image database, and loads the image data (DICOM header and encoded pixel data) into memory. The server then sends a precinct stream, also known as a JPP-stream, that contains a sequence of messages, with each message containing the data from a single packet. The client then parses the response stream, stores the packets in its cache, and renders the ROI. A cache model is maintained on the server side for each session, so the JPIP server can avoid sending redundant data to the client.
Figure 1. Architecture of the interactive-image communication system. PACS: picture archiving and communication system. JPIP: JPEG2000 Interactive Protocol. PDA: personal digital assistant.
We used a digital radiograph image with an original size of 2048×2500 pixels to demonstrate the interactive features of this image communication system. The JPIP client requests a low-resolution image to preview, and the data transmitted from the server is 2.13% of the full compressed image. With a low-resolution image to preview, the user can choose an ROI to view at higher resolution. Figure 2 shows the resulting images at a PACS image workstation. Using empty packets to reconstruct the tile data in order to raise transmission efficiency of the tile-part stream has been proposed previously.5 At the client side, we used a set of empty packets to reconstruct the code-stream so that the image information outside the ROI can also be decoded and displayed. This approach provides some navigation context in the regions outside the ROI.
In the case of browsing high-resolution medical images on a PDA, we used a DELL X51V Pocket PC connected to a wireless network. The embedded client software allows the PDA to communicate with the server. To demonstrate the protocol, we requested the previous image, this time to our PDA. First, we retrieved a low-resolution image to preview and then selected an ROI for higher resolution viewing. Figure 3 shows the resulting image on the PDA.
Figure 2. (a) The image is viewed at 25% resolution. (b) Spatial random access: Image reconstructed by the JPIP client at view-window: original file size = 2048×2500 pixels, requested file size = 512×512 pixels. The ROI in the red frame has full-resolution image quality, and the regions outside of the frame have lower resolution image quality.
Figure 3. Browsing a high-resolution image on a PDA: (a) Low-resolution image preview. (b) Reconstructed ROI image at view-window: original file size = 1024×1250 pixels, requested file size = 256×256 pixels.
We also tested the performance of the server under heavy load. It ran stably and reliably when processing large quantities of requests, and the average response time to each request was acceptable.
In our interactive-image communication system, the server works as the middleware between the clients and the PACS servers. This system allows both desktop clients and wireless mobile clients to efficiently browse high-resolution medical images. This system can also be expanded and integrated into regional healthcare systems, enabling medical information to be shared across wide-area networks.
This research was supported in part by the National Nature Science Foundation of China (Grant No. 30570512), the Department of Science and Technology of Shanghai (Grant No. 05DZ19510, 064119658, 06SN07111), and the Knowledge Innovation Engineering Plan of Chinese Academy of Sciences (Grant No. 07K2141208, 07K2142227).
Yuan Tian, Weihua Cai, Jianyong Sun, Jianguo Zhang
Laboratory for Medical Imaging Informatics
Shanghai Institute of Technical Physics
Yuan Tian is currently working toward her Ph.D. at the Shanghai Institute of Technical Physics, Chinese Academy of Sciences, China. Her current research interests include telemedicine, the JPEG2000 Interactive Protocol, and wireless medical systems.
Weihua Cai is currently working toward his Ph.D. at the Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai, China. His current research interests include image processing, JPEG2000 lossless and lossy compression for medical images, and computer-aided diagnosis.
Jianyong Sun received his Ph.D. in 2004 from the Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai, China. Currently, he is an associate professor at the Laboratory for Medical Imaging Informatics, Shanghai Institute of Technical Physics. His current research interests include teleradiology, image display, and visualization.
Jianguo Zhang received his Ph.D. from the Chinese Academy of Sciences, China, in 1991. From 1992 to 1994, he was an associate professor at the Changchun Institute of Optics, Chinese Academy of Sciences. From 1994 to 1998, he was a research fellow at the Laboratory for Radiology, University of California, San Francisco. Since 1998, he has been the director and professor of the Laboratory for Medical Imaging Informatics, Shanghai Institute of Technical Physics, Chinese Academy of Sciences. In 2004, he was appointed as a visiting professor at the Department of Radiology, University of Southern California. His current research interests include PACS technologies, medical image processing and visualization, imaging informatics, security technologies in health care, and electronic patient records. Dr. Zhang currently serves as an associate editor of the Chinese Journal of Computerized Medical Imaging and a panel member of the National Natural Science Foundation of China on biomedical engineering and biomedical imaging.