Application of computer-aided approaches to the PUMC classification of scoliosis

ABSTRACT

Surgical planning for scoliosis relies on the classification of the spinal curve pattern. To improve the reliability of the Peking Union Medical College (PUMC) classification system for scoliotic spinal curves, a computer-aided system is proposed and its reliability is evaluated. First, the reliability of curve measurements was improved by the computer-aided Cobb measurement approach. Second, judgmental errors were reduced by the computer program through the automation of the PUMC classification procedure. Four observers divided into an expert group and a resident group participated in the experiments. The kappa statistic was used to evaluate the variability. Classifications of 65 scoliotic cases by the four observers showed that with the computer's aid, the average intraobserver and interobserver kappa values were improved from 0.86 to 0.93 and from 0.75 to 0.86, respectively. The results indicate that the proposed computerized system can assist a surgeon in the PUMC classification of scoliosis and is especially useful for inexperienced surgeons.