The Visible Human Project
A Slice of Life
Washington Post Staff Writer
January 13, 1999; Page H1
The human body may be the world's most wonderful machine, but it differs from nearly every other mechanical gadget in one important way: It contains essentially no empty space. Built mostly of deformable substances, assembled in place and packaged in a custom-fit wrapper, everything in it fits perfectly.
This fact strikes home the first time a person sees a surgical operation. All of the body's functions arise from parts that are packed solid and working in total darkness. It's a revelation no less astonishing for being obvious.
People who want to work inside this universe are faced with an ironic truth. They must violate the body's natural order before they can understand or repair it. The first task of the anatomist and the surgeon is to create space and admit light. They can't work on the body's terms, no matter how good their skills.
This problem, of course, cannot be overcome. Nevertheless, for centuries, students of anatomy have sought ways to capture the body's complexity in its unperturbed state. Mostly, this has been done by drawing body components from dozens of angles and levels of dissection.
Pictures are made on the basis of structure (such as the skeleton), function (digestive tract), region (head or thorax) or layer (deep muscles of the back). These views, collected into atlases, are among the oldest, most important and most beautiful documents in medicine.
It's difficult to imagine that, after all this time, there's a better or even an untried way of depicting human anatomy. But there is, and it may turn out to be the best.
In 1993, the federal government commissioned a unique dissection. Two bodies, one of a man and one of a woman, were sliced into thousands of cross sections. Each cross section was photographed, and the images were transformed into digital code that can be processed and manipulated by computers.
In that form, the visual information became permanent and undecaying, at the same time capable of being dissected and selectively viewed without damage to the underlying whole. It was an anatomist's dream an undefiled body that could be explored from any angle or point of view, thanks to sophisticated software rendering images in three dimensions.
The idea for the Visible Human Project came from Michael Ackerman, a biomedical engineer at the National Library of Medicine (NLM), a branch of the National Institutes of Health and the largest medical library in the world.
In 1987, he gave a lecture at the University of Washington's medical school on use of computers in medicine. Afterwards, an anatomist suggested that, if the NLM really wanted to put computers to good use in medical education, it should find a way of computerizing an atlas of human anatomy.
Ackerman presented the idea to, Donald A.B. Lindberg, the NLM's director. Several years earlier, a long-range planning committee convened by Lindberg had suggested that the library, mostly a repository for words, try also to determine how to become a disseminator of images, which, after all, are at the heart of surgery, pathology and many other branches of medicine.
The library decided to pay for fine-sectioning and photographing two cadavers. Officials expressed hope that a few medical schools or "multimedia" companies might find a use for the data.
Today, more than 1,000 companies and institutions in more than 41 countries have been licensed to use information produced by the Visible Human Project for myriad purposes.
Several companies are making CD-ROM or Internet-based atlases that will allow students to virtually dissect a body layer by layer, isolate structures such as muscles and organs and rotate them in three-dimensional space. Other firms are using the data to make better crash-test dummies, design prosthetic joints and make simulators for common medical procedures.
In what may become the most clinically useful application, numerous companies are using the exquisitely detailed "data set" of the Visible Human cadavers to find ways to extract more information from the kinds of "imaging studies" made daily of thousands of people in hospitals or clinics.
Specifically, computer scientists and engineers are developing methods for turning the conventional two-dimensional cross-sectional images made by computed axial tomography (CAT) and magnetic resonance imaging (MRI) scanners into three-dimensional images that doctors can use to rehearse surgery and examine internal organs not easily or safely examined otherwise.
The computer's capacity to manipulate the data to obtain novel or custom-made views gives renewed meaning to the word "autopsy," which is Greek for "a seeing with one's own eyes."
"Some of the anatomical views were literally impossible before this," Lindberg says. "You even hear the experts say, 'Gosh, I've never seen such a view.'
The Visible Human's usefulness ultimately derives from the fact that it consists of lots of very sharp pictures.
Of course, the Visible Human data set doesn't have an infinite number of cross-sectional planes. But there are thousands, many more than had been recorded in a cross-sectional dissection of a human body.
Acquiring the images was a technical feat of the first order, and not one for the weak of stomach.
When the NLM advertised for bidders willing to cut a human body into thin slices and photograph each slice, there was far more interest than anyone anticipated. About 100 of the approximately 120 American medical schools, collaborating in various combinations, formed six consortia to apply for the contract.
The field was narrowed to three. Each consortium was asked to submit pictures of slices, taken 1 millimeter apart, from the abdomen of an animal or human body. One millimeter is about the thickness of a dime. A committee of anatomists and radiologists then chose the best. The winning bidder was a consortium of institutions in Colorado, Texas and Maryland.
The actual work on the cadavers first a male, then a female occurred at the University of Colorado Health Sciences Center in Denver. Under the direction of two scientists, Victor Spitzer and David Whitlock, the bodies first underwent head-to-toe CAT and MRI scans before being frozen solid.
To make them easier to handle, the bodies then were cut into four pieces with a special saw. Each of the three cuts separating the body into sections containing head, neck and thorax, abdomen and pelvis, thighs and knees, and legs, ankles and feet lost only 1.5 mm of tissue, represented by black space in the stacked final images.
It is more accurate to say that the cadavers were milled rather than sliced. The bodies were packed in dry ice and surrounded by a slurry of frozen alcohol at temperatures between minus 90 and minus 60 degrees Fahrenheit.
Starting from one end of a body section, a rotary rasp ground down the tissue to a specified depth. In the male, this depth was 1 mm. In the anatomically more complicated female, it was .33 mm, providing three times as much detail.
Each round of milling exposed a smooth, rock-hard surface in which the anatomical features were visible like the grain in a log. The surface then was photographed using both digital and conventional film cameras before removal of the next "cryosection."
Each cycle required from three to 15 minutes to complete, and the Colorado team could do about 50 each day. The work was meticulous because, once the rasp started, there was no second chance. The tissue came off in a frozen powder, which was collected, stored and ultimately "cremated in a respectful manner," according to an NLM official.
The Visible Human male comprises images from 1,829 cross sections and took nearly four months to complete. The female is made of images from 5,187 sections and took nearly 10 months. The NLM paid about $1.4 million for the work, which produced 15 gigabytes of computer code for the male and 40 gigabytes for the female. A gigabyte is 1 billion bytes, each of which contains eight digital units, or "bits."
But the result was essentially raw data when the library released it to the scientific community. Much of the work of the last four years, conducted in dozens of computer labs worldwide, has consisted of making the data more useable.
The digital data encode only what the cameras saw the surface appearances of each cut. The cameras didn't know what they were seeing. So, in effect, the entire data set must be "taught" anatomy before a person can hope to give it an order such as, "Show me the course of the sartorius muscle from where it originates on the pelvis to where it inserts on the tibia."
The job of delineating and labeling features is called "segmentation." Curiously, it's a step largely done by hand. Despite the high-tech nature of so much of the Visible Human Project, this essential task is performed in a way reminiscent of an artist's workshop of the late Middle Ages.
For example, Engineering Animation, Inc., of Ames, Iowa, won a contract from the NLM to segment images from the chest, or thoracic, cavity of the male. An anatomist would call up a cryosection on the computer screen and, using a mouse pointer, outline each anatomical structure. These delineated edges, and the names of the structures they outlined, then were added in digital form to the database. Medical illustrators reviewed the "hand-enhanced" images.
Engineering Animation and other companies are developing computer algorithms to mechanize, at least in part, the task of segmentation. Some programs can follow and label specific structures, such as blood vessels. But computers, in general, are not very good at detecting edges in images; humans are very good at it.
In addition, the images often include enough ambiguity to require decoding by a human eye. Fat can look like nerve tissue, and both, under certain circumstances, can look like bone. Only someone who knows what he or she is looking at can tell which is which.
"Developing generalized segmentation algorithms is still pretty far off," said Adrian Sannier, an Engineering Animation executive. Consequently, at one point, the company had 15 trained anatomists toiling at this work.
The project has given rise to numerous commercial products. Instructional aids for teaching anatomy are perhaps the most obvious uses, and many applications are being developed. It seems probable that computer-generated tours of the human body will become a standard part of anatomy courses.
Using the relationship between the highly detailed photos and less detailed CAT and MRI data, computer scientists are writing programs that turn the stacked, two-dimensional views of clinical imagery into a virtual three-dimensional reconstruction of an anatomical structure, such as a muscle or an organ.
The huge size of the data set, and the richness of the images, facilitates this difficult task. Using the 3-D cryosection reconstructions as a "gold standard" for quality, programmers can determine what information must be extracted from the CAT scans and MRIs in order to make clinically useful 3-D images from them.
"The importance of the Visible Human data set is that it unified the three modalities cryosection, CT [CAT] and MRI and that so much segmentation was done in one domain that could be applied to another," says Adrian Sannier of Engineering Animation.
Using lessons learned from the Visible Human data, the Iowa firm recently developed an algorithm for a company that makes "vascular stents."
Stents are hollow tubes inserted into diseased or blocked arteries to strengthen the vessels or restore their normal diameter. The company made stents to treat aneurysms, or a dangerous ballooning, of the abdominal aorta, the main-trunk artery that descends from the heart in the abdomen.
The stents are snaked into the aorta with a catheter, then sprung open in the weak spot, esssentially giving the vessel a new inner wall. Picking the right-sized stent is imperative.
The algorithm allows physicians to take normal CAT scan data from a patient and use it to construct a computerized 3-D model of the aorta, including crucial detail about internal contours and irregularities. This allows the pgysician to plan the procedure better and to choose the right-sized stent with more confidence.
Other companies are working on ways to explore other internal organs. Some experts predict that, in the future, procedures like colonoscopy now performed, uncomfortably, with a large fiber-optic device inserted into the colon will be done using conventional CAT scans reconfigured to produce 3-D images.
One of the most ambitious uses of the Visible Human data is underway at the Mayo Clinic in Rochester, Minn. Researchers in the laboratory of Richard Robb recently created a program that permits neurosurgeons to "see," on a video screen, what's ahead of them when they probe a patient's brain during an operation.
This is done in the operating room by placing the patient in an electromagnetic field that reveals the position of a special probe wielded by the surgeon. The probe's location is fed into a computer that orients it within stored images of the patient's brain made earlier by CAT and MRI scans. The software then generates a cross-sectional view of the brain that shows what structures the surgeon is approaching. The Visible Human Project data, of course, are not flawless. A few cryosections are marred by errors produced during freezing and milling. The NLM is paying several research groups to investigate these problems and suggest remedies. Also, there are regions of the body where even the .33 mm sectioning done on the female doesn't produce sufficient detail.
The head is the main place where that appears to be the case. The solution? Finer cuts.
Victor Spitzer, leader of the Colorado team, recently completed cryosectioning a human head and neck. He made 3,000 slices, one-third more than were done on the entire Visible Human male cadaver just five years ago.
© Copyright 1999 The Washington Post Company