With standard 10Based-T network, the user starts seeing some interruptions that interfere with the navigation. While the animation speed is greatly reduced, the study found that the low bandwidth still gives the user acceptable interactivity. This is analogous to reading data from a slow CD-ROM reader rather than a hard disk. Technological advances and competition are replacing 10Base-T networks and slow CD-ROM readers with economical and faster alternatives. Supporting lower resolution images (in addition to the higher resolution version) for network users also helps solve the problem of network speed and uncertain loading time.

      As a client-server application over a local area network, a simple notebook PC can be linked to a high-end graphics workstation or a supercomputer. From a conference hall, a lecture theatre, or a meeting room where you might want to show the results of your work, this setup makes the scientific results easily portable and allows you to share your last-minute information with your audience. There is no need to create videotape or lossy-compressed video and in the process suffer degradation in quality and interactivity.

      With a link like this between a PC and a supercomputer, the PC and the image browser software become an intelligent display and front-end for the computational power behind it. Searching the full Visible Human Project data set for the range of images to load for processing, displaying the results and doing interactive quality control become easier and faster. The data to load may not be the images themselves but the numerical data of X-ray, or nuclear sensitivity for modeling safe dosages, or any similarly numerically intensive task. Many of the image processing tasks, previously considered to be in the workstation computer domain, can of course now be done directly on the PC itself. In fact, image processing can go on during the animation.

      The link may be continuous, as in haptic simulations of needle insertion. The fast display capabilities of the PC mean that the appropriate image at the point of the needle may be directly accessed from disk and workstation memory does not need to be set aside for this. The basic core modules of the image browser allowing fast access to the disk is separated from any single user interface so that they can quickly be implemented with other links to the user or to another computer.

      With the Visible Human Project data set itself, consider a medical university setting with a central site of a full-resolution database and display PC's distributed among the students. The central data set may be augmented with full voxel classification and linkages to other data. Each student with the lower level Visible Human Explorer CD-ROM on his local machine is able to scan through the entire Visible Man with linkage to the 1500 anatomical structures. This version has full spatial resolution images but at 15 to 30 mm spacing rather than 3 mm between images. This is sufficient for many learning situations. However, the system would be set up so that once the student stopped scanning through the local images and stopped at a single image, full-resolution images on either side would be automatically or on request downloaded from the central site. Similarly, ancillary data would be interactively available, linked to mouse clicks on the local image.

      For network applications a tradeoff can be made between bandwidth and animation speed. As more images per second flash by the user, they can be of lower resolution. Once the user slows down the number of images per second, he or she will be able to catch finer detail and the image resolution can be increased to make full use of available bandwidth.

      Now the doctor with the unusual case may be on the network, not at the central site. What about doctors at local rural hospitals however? The omnipresent Internet becomes a consideration.