Visible Human Virtual Colonoscopy

Lichan Hong, Arie Kaufman,

Zhengrong Liang, Ajay Viswambharan, and Mark Wax

State University of New York at Stony Brook Stony Brook, NY 11794

Colonoscopy and barium enema are the only means currently available for detecting polyps of the human colon. Unfortunately, colonoscopy, an expensive invasive procedure, requires intravenous sedation and takes approximately 45 minutes to perform. On the other hand, barium enema, although less expensive, requires a similar length of performing time, a good deal of patient positioning and thus patient cooperation. Results of barium enema vary according to the technical quality of the procedure and the experience of the interpreter [1]. Colon polyps that are 5 mm or larger are considered clinically significant since they have a high probability of being malignant.

Recently, we have proposed an alternative procedure called 3D virtual colonoscopy [2], which utilizes computer graphics technology to visualize the inner surface of the colon based on CT images of the abdomen, and therefore assists the physician in detecting the presence and characteristics of the abnormalities inside the colon in a non-invasive manner. A set of newly developed visualization algorithms [2, 3] has been applied on acquired CT data of pipe phantoms (made from respirator tubes coiled in water tanks containing several 3 to 7 mm rubber cylinders simulating polyps) and patients, as well as the Visible Male data. Our algorithms support both fully-automated fly-throughs and interactive walk-throughs along the inside of the colon. To generate a fly-through automatically, we employed an algorithm which computes a flight path while guaranteeing that the virtual camera always stays in the center of the colon to achieve better viewing of the colonic surface. For interactive navigations, we have also designed a physically-based camera control model which uses the concept of potential field to guide the physician through the colon while allowing him or her to study a specific lesion in detail. Combined with a newly proposed Z-buffer assisted visibility algorithm, high frame rate (approximately 10 frames/sec) navigations were produced on a SGI Onyx graphics workstation.

When testing our algorithms on the Visible Male data, we primarily focused on the physical cross-sections of the data set, from slice No. 1595 to slice No. 1848. We chose to use the 24-bit cross-sections mainly because the difference between the colonic surface and the residuals remaining inside the colon was easier to perceive than the corresponding CT and MRI data. In addition, the color information maintained in the 24-bit cross-sections has been utilized to achieve realistic visualization. Due to memory limitations, each slice of the physical cross-sections was down-sampled from the original resolution of 2048 x 1216 to 683 x 406. The red component of the 24-bit color data set was extracted and manually segmented, and subsequently reconstructed into a 3D representation and visualized with our algorithms. When rendering time was not a major concern (i.e., in the automated navigations), the whole 24-bit color data set was used as a texture map to generate the animation frames. Figure 1 shows three images from an automated fly-through of the colon of the Visible Male. Please check out our page at

    
http://www.cs.sunysb.edu/~volvis/colonoscopy/colonoscopy.html

Our initial work on 3D virtual colonoscopy has demonstrated that a cost-effective, non-invasive, patient-comfortable, and efficient procedure for detection of colon polyps at the size larger than 3 mm in diameter is feasible. For the Visible Human data, we are currently designing a semi-automatic segmentation algorithm and also planing to apply our visualization algorithms on the Visible Female data set as well.

   (a)              (b)                  (c)

Figure 1: Three images from an automated fly-through inside the colon of the Visible Male.

References

[1] C. Morosi, G. Ballardini and P. Pisani (1991). "Diagnostic Accuracy of the Double-Contrast Enema for Colonic Polyps in Patients with or without Diverticular Disease". Gastrointest Radiology, vol. 16, 346-347.

[2] L. Hong, A. Kaufman, Y. Wei, A. Viswambharan, M. Wax and Z. Liang (1995). "3D Virtual Colonoscopy". Proc. IEEE Biomed Visualization Symposium, IEEE CS Press, Los Alamitos, CA, 26-32.

[3] L. Hong, S. Muraki, T. He and A. Kaufman(1996). "Physically-Based Interactive Navigation". Technical Report TR.96.01.09, Computer Science Department, SUNY at Stony Brook, 1996. Also submitted for publication.