Rheumatoid arthritis (RA) is a very painful condition with tremendous societal impact. Nearly one percent of the population suffers from RA and the annual cost to the North American healthcare system from arthritis in general has been estimated at $64 billion. Symptoms range from mild discomfort and pain to loss of joint function as the disease progresses to its end-stage. This enormous healthcare problem is best met by the prescription and development of effective therapies. In order to evaluate these therapies, highly accurate and reproducible methods are required to quantify the state of the disease. Radiographic evaluation of hand films is currently used to assess disease progression though the use of semi-quantitative subjective scoring systems. These methods, however, are subjective and suffer from significant reader variation. In addition, the need for specialized training makes the systems costly and difficult to implement on a widespread basis. There is currently no truly quantitative method to assess arthritis progression in the affected joints. To address this need we propose to apply sophisticated image processing, multivariate analysis, neural networks, and regression tree methods to hand radiography. We will perform a quantitative and systematic study of radiographically visible structural changes due to RA. This work will provide previously unavailable objective and disease sensitive radiographic outcome measures of RA progression. The result will be a computer-based system with improved disease sensitivity, which will lead to more accurate evaluation and appropriate prescription of therapies. This work will play a major role towards alleviating the effects of this debilitating disease.
Showing the most recent 10 out of 11 publications
