by Karen Moltenbrey
Throughout the centuries, tomb raiders have destroyed countless fragile Egyptian mummies while ravaging ancient pyramids and crypts in search of treasure. Even archeologists unwittingly contributed to the de struction of these preserved bodies. Their invasive re search methods-which often involved the removal of elaborate casings and the unwrapping of bandages-were stopped decades ago when it was discovered that these practices were causing rapid decay of the mummified remains.
As an alternative research method, the Man ches ter Museum in England, which has a large collection of Egyptian mummies, has been pioneering noninvasive testing to advance its research without opening the mummies' bandages. For some time, the mu se um has been using X-ray images and, more recently, computed tomography (CT) scans as visual tools for studying the bones and tissues of the mummies in its collection. Both these techniques highlight hard, dense matter such as bone and gold (found in the funerary decorations of the mummies). In particular, the museum has used CT scans to guide fiber-optic tools through the bandages to extract tissue samples from one of the mummies. However, using this type of imaging technology to "see" what lies beneath the exterior layers of its subject presents some difficulties for the museum researchers.
|The Manchester Visualization Centre used a unique volume-visualization technique to help a museum further its noninvasive research for studying mummies.|
With CT, used primarily in medical applications, a series of 2D "slices" is acquired from a 3D object that can then be stacked in a computer to digitally re-create the 3D volume. The advantage of CT scanning for viewing certain medical data is that it provides more information than a conventional 2D projection produced by an X-ray and can highlight objects with varying densities-for instance, a tumor growing among tissue or bone. However, CT scans do not allow a person to see inside the mass, or volume. As a re sult, interpreting the scanned data of a mummy becomes especially challenging, as the mummification process has drastically altered the subject's physiology, making it extremely difficult to distinguish between the densities of the soft tissues and the bandages wrapped around the body.
Now the museum has found a way to further advance the study of mummies, this time through its pioneering use of X-ray/CT imaging and new 3D computer reconstruction techniques for more robust viewing of 3D volumes. Working with the Manchester Visualization Centre, the museum's Egyptology Department is using Advanced Visual Systems' (Waltham, MA) AVS/Express software in conjunction with an innovative volume-rendering approach to examine the interior structure of mummies without inflicting irreparable damage.
With volume rendering, visual images are displayed directly from the acquired volume data by assigning values to each of the 3D coordinates (called voxels, or volume elements) in the data set. With this type of volume visualization, the viewer can see the internal structure of the 3D data, differentiating it from conventional surface-based representations in which polygons are placed on flat wireframes that can be twisted and rotated but can never truly simulate the interior of the 3D object.
For the museum study, the Manchester Visualization Centre adapted the new volume-visualization method used in its med i cal research for placing cochlea im plants in the inner ear to correct hearing impairments. As with the mummy application, the 3D volume-rendered images of the inner ear are easier to examine and un der stand than the individual 2D slices of a CT (or an MRI, in this case).
Because of the Manchester Visual i za tion Centre's experience with using volume vi sual ization techniques in the medical field, the museum approached the organization, "to see if we could do anything with the CT scans of a particular mummy that would help the museum researchers take a better look inside," recalls Joanna Leng, visualization support officer at the Centre.
Of particular interest to the museum was obtaining further data about an adult female mummy, known simply as Mummy 1766, who lived during the first-century AD. The museum chose this mummy because she had not been damaged either during her recovery from Egypt or by earlier research.
Using standard volume-visualization techniques (such as CT scans) seemed like an ideal way to view the mummy's interior when the researchers initially tried them, but Mummy 1766 presented unique difficulties. In traditional volume-visualization applications, the transparency of each data point in the volume is linked to a data value, so that one type of feature-for instance, hard matter such as bone-can be highlighted. This works especially well for medical-scan data because there is a significant contrast in the data (bones versus tissue). But the mummy's body is tightly wrapped in multiple layers of resin-coated bandages painted with funerary scenes, so that it was difficult to distinguish the soft-tissue data values from those of the bandages. Further affecting the CT imagery were the complex and distorted shapes of the mummy's soft tissues, caused by dehydration.
A gilded cartonnage cover made of three separate pieces-for the head, breast, and feet-also contributed to incomplete CT images. "The gold is about the same density as the bone, so it masked the skull," explains Leng. Complicating the data further was the cover's complex shape, which in the case of the gold mask, obscured the front and sides of the skull and face.
|A gilded cartonnage cover made it difficult to use standard volume-visualization techniques on the mummy because the densities of the gold and the bone were too similar.|
At first, Leng considered using a surface-based technique developed in Germany for a similar application of displaying the components of a mummy. Through surface rendering, surface or line objects with the same data value, called isosurfaces, are extracted from a complete 3D data set and converted to geometric primitives for viewing. However, Leng believed that the 47 layers of bandages around Mummy 1766 would present a problem in this instance. Not only would it have taken considerable time to place isosurfaces around all the data points for each bandage layer, but the small differences in the data values for the bandages and preserved soft tissue would have made it difficult to select an appropriate isosurface level.
"A surface is a clear boundary between objects, but [in this in stance] no clear boundary exists. So if we were to have used this technique, some areas of bandage might have been confused with soft tissue, resulting in the wrong surface," explains Leng.
Leng also dismissed using two other types of conventional rendering techniques: active contouring and region growing. With active contouring, the user isolates, or segments, a slice of the data set by drawing a rough contour around an object. Similar contours from other data slices are then added until the 3D model is complete. With region growing, the user selects a data point from the middle of an object, and a computer program iteratively adds similar neighboring data points according to criteria preset by the user until the entire object is selected.
The problems, according to Leng, are that active contouring would have taken too long for this project, given the Centre's time frame, while region growing often results in inaccurate data if the preset criteria are too rigid or too lax. Furthermore, the data resulting from these approaches is segmented and, therefore, is surface-rendered instead of volume-rendered.
Instead, Leng used a combination approach of region growing and active contouring that maximized their effectiveness while overcoming their individual inadequacies-and enabled the data to be volume-rendered.
First, Leng scripted a series of computer programs in C language running on an SGI (Mountain View, CA) O2 workstation that segmented and manipulated the CT scan of Mummy 1766. This region-growing algorithm took a user-defined point close to the back, central part of the mask for each 2D data slice and grew it, through active contouring, as a line in both directions along the back of the mask and off the edge of the slice. Most important, she notes, the contour was not used for placing a surface over the object (as it is for region growing and active contouring, where the contours become part of the surface). Rather, the contour was used to divide the data, so the data values on all sides of the object were preserved by segmenting compound objects instead of single objects. Leng then separated the data values on each side of the contour with an algorithm so that the separated data could be volume-rendered without affecting the other data.
This process enabled Leng to create three separate data sets: one of the mask, one of the mummy's head with its bandages, and one that was a combination of the head and mask. She then manipulated the data set so the mask data no longer obscured the head data.
Using AVS/Express, a visualization software package with some volume-visualization functions, Leng volume-rendered the data sets, which resulted in detailed images of the layers, including the mummy's body beneath the bandages. Unlike surface rendering, volume rendering allows details-no matter how small, ill-defined, or discontinuous-to be drawn from the data by altering the transparency of the data points that make up the surface of an object. This enabled the Egyptologists at the Manchester Museum not only to visualize the bone and soft tissue, but to see exactly where that tissue met the bone. "As a result, you get a more realistic idea of what the mummy looks like," notes Leng.
In essence, Leng's volume-visualization technique enabled the researchers to remove the cover from the mummy's head, then the bandages from her face, and the soft tissues from her skull. "This gave us an idea of where the mask was placed in relationship to the face. When you look at one or the other, you have an idea of their shapes but not how they line up. So you couldn't tell where her chin was under the mask," she says. The 3D imagery also revealed that the mummy's neck and head are twisted under the mask-a fact that was previously unappreciated but may aid further research.
|To separate the scanned cover data from the skull data, the Visualization Centre used a combination of region growing and active contouring, two types of rendering technologies. This let the researchers "lift" the mask so it no longer obscured the head. |
"When I showed the imagery to Rosalie David, professor of Egyptology at the museum, she said it was as if she were seeing this mummy, which has been in bandages for 2000 years, suddenly un wrapped," Leng explains. "Using this technology, we have been able to uncover some of the secrets of the mummy without destroying its exterior, such as the perfect condition of her skull and teeth. Pre vi ously, you would have had to unwrap her to get an idea of her body's condition."
Prior to these latest studies, re searchers had theorized that Mummy 1766 perhaps died of Schis to soma, a blood disorder still prevalent in many third-world countries. To further investigate that theory, the museum researchers had used CT scans as a visual tool to help them extract tis sue sam ples that could be compared with pres ent-day human tissue, with the eventual goal of developing a biochemical test for the disease. Using this new volume-rendering technique, the re search ers could more quickly and accurately ex tract such a sample, causing as little harm as possible to the mummy.
"The museum is interested in how the Egyptians lived thousands of years ago, what types of diseases they encountered, and what kind of diet they had," Leng says. "So its goal is similar to that of doctors using medical visualization today-to obtain as much information noninvasively as possible."
Karen Moltenbrey is an associate editor of Computer Graphics World.