REPETITIVE motion injuries are one of the fastest growing causes of lost time to business and industry, not to mention their impact on worker health and morale. The orthopedic surgery these injuries sometimes necessitate is costly and painful. The therapy and orthopedic implants associated with degenerative bone and muscle diseases and acute injuries are also costly, and in the case of implants, the initial cost may not be the final cost if and when the implant needs to be replaced. Computational models of joint anatomy and function can help doctors and physical therapists understand trauma from repetitive stress, degenerative diseases such as osteoarthritis, and acute injuries. Models of prosthetic joint implants can provide surgeons and biomechanical engineers with the analytical tools to improve the life-span of implants and increase patient comfort.1 With such purposes in mind, the Laboratory embarked about three years ago on a mission to model the whole human hand at high resolution. The challenge is that most biological structures are dauntingly complex, and the hand is no exception. The human wrist alone has eight bones, and the rest of the hand has 19 more, to say nothing of soft tissues--ligaments, tendons, muscles, and nerves--and the interactions among them. More recently, the Laboratory's Computational Biomechanics Group within the Institute for Scientific Computing Research (ISCR) narrowed the mission to a computational model focusing on the dynamics of specific bones and joints that are often associated with injury or damage. The group also undertook a closely related endeavor: creating a computational model of prosthetic joint implants, initially for the thumb. In light of the complexity of these models and the need for very high accuracy, it is appropriate that a facility like LLNL--which offers powerful computational resources, an understanding of complex engineering systems, and multidisciplinary expertise--take on these tasks. It is also significant that the work is being done collaboratively through the ISCR and draws on experts from the Laboratory (particularly the Mechanical Engineering Department), academia, medicine, and industry (see the box below). The NIKE3D modeling code, for example, which was developed at the Laboratory to address engineering problems involving dynamic deformations, such as the response of bridges to large earthquakes,2 is now being used as part of our collaborative joint modeling work. Each person's bones differ in shape and size. Our models are based on the detailed anatomy of individual people. We start with high-resolution data obtained from computed tomography or magnetic resonance imaging, as shown in the illustrations below. Images from a single hand scan involve several gigabytes of raw data, and the models developed from them are highly complex--thus the need for powerful computers. Focusing on the Hand and Knee We focused our initial attention on a few joints in the hand. One joint of considerable clinical interest is the thumb carpo-metacarpal (CMC) joint, which connects the long bone at the base of the thumb with the wrist. During routine grasping activities, CMC joint surfaces are subjected to total forces greater than 200 kilograms (440 pounds), so it is not surprising that injuries are common. The thumb is also often involved in repetitive motion injuries, and the CMC joint is the structure most affected in osteoarthritis, which strikes 8% of the U.S. population. Other joints of considerable interest are the knee and the proximal interphalangeal joint and the metacarpo-phalangeal joint in the index finger, which have some of the strongest ligaments in the hand. Previous analyses of joint function (for example, rigid-body kinematic analyses) have typically provided less information than is possible through finite-element methods. On the other hand, most finite-element analyses of biological systems have been linear and two dimensional. We are applying three-dimensional, nonlinear, finite-element codes that assign material properties to bone and the soft-tissue structures associated with joints. For example, using the NIKE3D modeling code to look at biological problems for the first time, we can model bones as materials that are more rigid than tendons, but less rigid than a metal implant. Soft tissues are inhomogeneous, undergo deformation, and some show elastic behavior. Our methods simulate tissue behavior under various loads, and joint movement with and without prosthetic implants, and they can assess injury following trauma, such as that caused by a car crash. Because we can assign different material properties to the tissues and examine a range of loads that are experienced in real life, our methods allow us to see interactions between tissue types for the first time, and we can identify regions of high stress. Our high-quality visualizations, such as the one shown in the illustration above, display complex results in easy-to-understand form. Two main issues are the computer time needed to run a finite-element code and the labor needed to develop a model from each new scanned data set. We are addressing the first problem by working on powerful supercomputers and by preliminary research in partitioning biological finite-element models to run on massively parallel machines.3 To address the second problem, we are automating as much of the modeling process as possible.

Other 3D Visualization Tools
We are also designing three-dimensional computational tools to interactively move tissues so we can simulate surgical procedures. In the future, we plan to extend our modeling to include additional joints and perhaps internal organs, such as the heart, lungs, and liver.
How can our computational tool benefit the clinical community and ordinary individuals? Our models provide data on internal joint stresses and strains that are not otherwise obtainable. They can be used by surgeons to help plan treatment and to assess outcome following a traumatic or repetitive motion injury. They can help a surgeon predict results, such as strength, range of motion, and other indicators of function after an operation.
Orthopedic implants are a multibillion-dollar U.S. and worldwide industry; however, today's prosthetic joint implants have high failure rates. They often loosen, wear, and fail before the end of a recipient's life, necessitating painful and costly replacement. Orthopedic implants last on average 5 to 15 years. The initial implant can cost about $20,000; revision implants are more expensive. Our methods can eventually be applied to any human joint for which prosthetic implants have been designed. Our models are leading to better designs for prosthetic implants, resulting in longer life spans and fewer costly followup operations. Finally, our work can help the automobile industry to develop safety features that will protect against injury to the head, chest, and lower extremities.

Collaborators in Biomechanics Modeling
ISCR biomechanics research is collaborative in the broadest sense. At Livermore, we work with experts in computer vision, mechanical and electrical engineering, nondestructive evaluation, health care technology, health services, and with visiting scholars and students. Partners outside the Laboratory include:

Academic institutions
University of California, Berkeley
University of California, San Francisco
University of California, Davis
University of California, Santa Cruz
University of New Mexico
Institute for Math and Computer Science, Hamburg, Germany

Medical facilities
Kaiser Permanente
G. W. Long Hansen's Disease Center
Louisiana State University Medical Center
Massachusetts General Hospital
Children's Hospital, San Diego

Industry
ArthroMotion/Avanta Orthopedics, Inc.
ExacTech
Orthopedic Biomechanics Institute
National Highway Traffic Safety Administration
Wright Medical, Inc.
XYZ Scientific Applications, Inc.
Zimmer, Inc.

Key Words: biomechanical modeling, finite-element modeling, Institute for Scientific Computing Research (ISCR), NIKE3D, prosthetic point implants.

References
1. Modeling the Biomechanics of Human Joints and Prosthetic Implants, UCRL-TB-118601 Rev. 1, Lawrence Livermore National Laboratory, Livermore, CA (1995).
2. Energy & Technology Review, UCRL-52000-95-9/10 (September/October 1995) is devoted to a series of articles on computational mechanical modeling, including NIKE3D.
3. For more information on finite-element modeling using massively parallel processors, see "Frontiers of Research in Advanced Computations," Science & Technology Review, UCRL-52000-96-7 (July 1996), pp. 4-11.