Abstract—The purpose of this study is to develop a computer-aided decision making system for scoliosis treatment based on biplanar spinal radiographs. Three-dimensional (3D) features were automatically extracted from the 3D spinal model that was reconstructed by using the self-calibration algorithm from landmarks identified by users on radiographs. The k-nearest-neighbor model was trained and then was used to determine the treatment for a scoliotic curve as observation, bracing, or surgery. With leave-one-out methodology, 31 cases were used to test the system performance. Experimental results showed that the system could achieve accuracy of 91.9% and consistency of 96.8%. This system can be an objective aid to surgeons in the task of treatment decision for scoliosis.  
 
Index Terms—Idiopathic scoliosis, treatment decision, radiograph, three-dimensional feature

Cite: Junhua Zhang, Fei Guo, Hongjian Li, and Liang Lv, "Computer-Aided Treatment Decision on Scoliosis Based on Three-Dimensional Radiographic Features," Journal of Medical and Bioengineering, Vol. 3, No. 3, pp. 212-216, September 2014. Doi: 10.12720/jomb.3.3.212-216