Computer-assisted Cobb angle measurement on posteroanterior radiographs

Abstract

The Cobb angle method is the gold standard to assess severity of scoliosis. A computer-aided method was developed to provide a semi-automatic Cobb angle measurement during a scoliosis clinic. This study was to evaluate the reliability and accuracy of the developed method. Curve types were also tested. The computer method required enhancement of the contrast, normalization of the image size, and selection of the end-vertebrae on the radiographs before the automatic measurement started. The computer-aided process automatically identified the line segments that fitted to the endplates of the end-vertebrae. The Cobb angle was then calculated from the slopes of these lines. Seventy-six radiographs were randomly selected and categorized with Lenke's classification. Among them, 75 cases were used and categorized into 4 types: 1, 3, 5 and 6. One type 2 case was excluded. An orthopedic spine surgeon measured the radiographs manually, serving as the reference standard. Two observers used the developed method and measured twice. For each curve type, the inter-method, inter-observer, and intra-observer variability were analyzed by Intraclass correlation coefficients (ICC[2,1]). The ICC values were higher than 0.90 in all these types. The developed method was reliable to measure the Cobb angle and was not dependent on the curve type.