Web Magazine for Information Professionals

OMNI-Corner: Meeting the Visible Human

Sue Welsh reports from the Visible Human Project Conference of October 1996, an event that brought together many of the people involved with one of the most high profile Internet-based medical resources.

The Visible Human is one of the Internets best known sites; one of its big stories. However, the size of the Visible Human Dataset (VHD) means that most of us have no way of making use of the raw data itself. We have, so far, been more impressed by the idea than the reality.

Not for much longer - last month the first Visible Human Project Conference brought together the people involved in producing the data, the researchers, information scientists and clinicians who have used it and the ignorant but curious, for two days of demonstration and discussion, at the National Institutes of Health, Bethesda, Maryland. What emerges is not a white elephant but a resource being used by many in research, medical education and, especially, in commercial products.

The Origin of the Species

The Visible Human Project was first proposed in the late 80s, when the tools for really effective dissemination of such a dataset did not exist or were in their infancy. The foresight of the NLM, therefore, is to be admired. As early as 1986, the NLMs long range planning foresaw a coming era where NLMs bibliographic and factual database services would be complimented by libraries of digital images, distributed over high speed computer networks..... [1]. In 1989 the Board of Regents approved the principle and further committees made the decision to create a digital image library based on two cadavers, on male, one female.

The male dataset was the first to appear. Consisting of MRI, CT and anatomical images at 4mm, 1mm and 1mm respectively, the total dataset is 15 gigabytes. These simple facts do adequately tell the story of the efforts involved in the acquisition of the data, which was clearly conveyed by Victor Spitzer on the first morning of the conference. The anatomical images were produced, as everyone knows, by literally slicing up the cadaver. There are 1871 cross- sectional anatomical images; on average, each took 6 minutes to make. To make the cadaver more manageable it was sawed into four pieces before the slicing began, with some loss at the junctions. Internal cavities appear blue on the anatomical images, as they were filled with latex. The cadavers arms were glued to its side to preserve their relation to the trunk from one image to the next; similarly the cadavers legs were fixed at the ankle. The cadaver had to be squashed to fit into the MRI scanner, and so these images do not coincide with the CT and anatomical data.

The female dataset was produced in essentially the same way, but its bigger - about 40 gigabytes - as the anatomical slices were taken at more frequent intervals, there are 5,000 slices in all.

The NLM makes this resource available without charge across the Internet, for those satisfying their license agreement [2], and essentially, the Visible Human Project Conference was a showcase for people whove taken them up on the offer. Projects from around the world brought their software to the Natcher Conference Building in Bethesda, Maryland. Silicon Graphics supplied the hardware, and the show began!

Meeting the Visible Man

For most of us, a visit to the Visible Human web site is a two-dimensional experience. Held back by our browser, we stroll through the dataset and are presented with a succession of flat, anatomical, CT or MRI images, similar to flicking though a standard cross-sectional anatomy text-book. This approach is one valuable use of the data, and some projects were at the conference to demonstrate viewers for the data as it is, but to comprehend the real potential of the resource, its necessary to think in 3D.

Many of the projects demonstrating their expertise here are using the VHD to create 3D images of the human body and its internal structures. Consequently, we were treated to so many fly- pasts and fly-throughs that it felt like being an extra in Toy Story. Seeing a camera traveling up the spinal column, around the inside of the respiratory system and though the veins and arteries is an unforgettable experience. How is it done?

Creating three-dimensional structures (volumes) out of 2 dimensional data is essentially a two step process. Firstly, the boundary of the anatomical object must be identified in the 2D images, a process known as segmentation. Secondly, the segmentation data must be interpreted in three dimensions, or rendered. Once this is achieved natural looking colours can be artificially added, or false colours can be used to highlight specific structures.

Any identifiable structure or surface can, in theory, be thrown into 3D this way. Perhaps the most extraordinary images from this conference were not the moving pictures of the inside of the human body, although these are striking, but the projects that included 3D images of the skin, thus showing the audience what the Visible Human people looked like when they were alive.

Among the many examples of these processes presenting their findings at the Conference were:

The next step forward, once you have generated your 3D images, is to make them behave like the human body does. Virtual reality beckons, in the future medical students will be able to practice procedures on realistic, computer generated models, before they are allowed access to patients, the benefits are obvious.

To achieve this, small volume units, or voxels, in the 3D volume must be associated with physiological properties, and programmed with the basic principles of physics, so that they respond appropriately to interference from outside. One project described how they we using the VHD and high tech interface devices to replace the orange as a model of the human body for practicing injections, another suggested that this arrangement should be turned on its head, that physicians should demonstrate procedures on simulated patients, and the computer should then perform it.

As well as its obvious uses in medicine, academia and research, the VHD has been pounced upon by many commercial companies, who have not been slow to recognise the market for these exciting images. Although the dataset itself is huge, and the manipulation of it requires computing power beyond the reach of the average pocket, the products of the 3D visualisation process fit nicely into CD-ROM and can be viewed with a standard desktop computers.

Engineering Animation Inc. (Mosby) were most in evidence, demonstrating several CD-ROM products, including the Dissectable Human, which allows you to peel back the Visible Human skin, to see the musculature, the muscles to see the skeleton, view the nervous, digestive and respiratory systems etc. Visit their site to see the sort of educational material which can be created from the Visible Human. [3]

Future Directions

This was not the conference for discussion of the future of the project, with so much to be seen which is being done with the data as it is at present. We learned, however, that the National Library of Medicine has plans to link the image data to other knowledge based, and that research is underway to achieve the connection of structural-anatomical to functional- physiological-knowledge [1] with the goal of bringing together the print library and the image library in a single resource.

Virtual anatomy is already here and virtual physiology around the corner, where will we go next? One delegate was for virtual biochemistry! Another, perhaps more feasible in the short term, for a increase in the population of visible humans, with new projects providing data for different races and ages. Whatever the NLM decides, it seems that the Visible Humans have a great future ahead.

[1] The Visible Human Project (factsheet) : http://www.nlm.nih.gov/publications/factsheets/visible_human.html

[2] Agreement for use of images from the Visible Human Dataset : gopher://gopher.nlm.nih.gov/00/visible/vhpagree.txt

[3] The Dissectable Human (TM) CD-ROM : http://www.eai.com/interactive/dhuman/dhuman_cdrom.html