UCSF manages a testbed network linking UCSF, Mt. Zion Hospital, and SF VA Medical Center, allowing neuro-images to be transmitted in real-time to experts at UCSF. Physicians can access the large-scale image and related data file server from their desktops. The group is also designing an integrated cardiac care unit (CCU), combining multi-source patient information for distribution to nurse stations and patients' bedside monitors. In addition, UCSF will soon be connected to NLM via the NASA ACTS Satellite, enabling access to the National Health and Nutrition Examination Surveys and the 3-D Visible Human image databases.
Each of these projects need to extend wireline connectivity to bring medical image, database, and patient care information to mobile medical professionals, in the "unwired" parts of the hospital, e.g., nurses stations, patient wards, operating rooms, and emergency rooms. The applications support technology we will develop will be stressed to provide adequate performance for high resolution image transmission over modest bandwidth wireless links.
We are developing mobile retrieval applications for medical image transmission that exploit our application support technology for agent-based compression and progressive layering of data types. For example, medical images from the UCSF PACS system require from 0.8 to 4 Mbps, and are within the bandwidth range of in-building wireless networks if lossless compression is used. More aggressive strategies for image compression and layering will be pursued for transmission outside the building. For example, ultrasound images are inherently noisy and can tolerate increased compression without serious loss of information. Furthermore, areas of rapid change, such as the movement of the valves in a beating heart, contain inherently more information than the areas of the image that are not changing as rapidly. This can be exploited to achieve much reduced bandwidth demands. In the course of the research, we are developing a range of such mechanisms to support a rich collection of medical image and patient information data types.