Virtual Human Anatomy and Surgery System

Abstract

Historically, medical students have practiced on cadavers to learn human anatomy, as have physicians wanting to brush up on their knowledge. However, because of storage cost and limited availability of cadavers, practice on cadavers has proven problematic. As computers become more powerful, medical professors have dreamed of a day when they will be able to dissect bodies with the assistance of virtual reality. We have developed the Virtual Human Anatomy and Surgery System (VHASS) as a potential solution. VHASS uses cryosection images (natural-color images generated by slicing a frozen cadaver) to reconstruct computerized three-dimensional cadavers. VHASS enhances human anatomy education by creating three-dimensional volume models that include details of human organs, giving medical students and physicians unlimited access to realistic virtual cadavers. Major components in VHASS include three-dimensional virtual humans, direct volume rendering of virtual humans, surface models of segmented human parts, and real-time manipulation on virtual humans. Direct volume rendering on un-segmented cryosection images is still an open research topic. Different from traditional volume rendering, which uses transfer functions to map scalar values to colors and opacity, direct volume rendering on cryosection images needs efficient transfer functions mapping vectors to opacity, which is complicated by the non-linearity of color space. We have created a series of new transfer functions for volume rendering on un-segmented cryosection images. To create human part surface models, we separate human tissues within cryosection images, dissect all human organs according to their anatomic structures, and reconstruct a three-dimensional volume model for each part. VHASS renders each part as a high-resolution, natural-appearance three-dimensional model and labels it properly to facilitate learning. This enables users to group different parts to better understand human anatomy. VHASS allows real-time interactions, such as drilling, scanning and slicing on human parts. We re-generate human part surface models at run-time for deforming interactions. We have analyzed the limitation of the well-known Marching Cubes algorithm and modified the algorithm to work with our data. We also have developed a new neighbor-based surface reconstruction algorithm, which has the same performance as the Marching Cubes algorithm but without the limitation of the Marching Cubes method. For better performance, the new algorithm has been ported onto the new graphics hardware using the geometry shader. Our implementation on the geometry shader serves as an example of exploiting the new GPU parallel processing hardware. VHASS supports stereo rendering, haptic interaction, tracking and three-dimensional content production. Using the Sharp three-dimensional display on a laptop, VHASS provides low-cost, portable stereo rendering of human parts without the requirement of special glasses. Integrating with large size stereo projector and ultrasonic trackers, VHASS allows people to manipulate human parts in the immersive stereo environment. By integrating SensAble Onmi haptic device, VHASS enables people to feel the touch on human parts. VHASS integrates three-dimensional content creation by allowing students to print out physical models of human parts.