Automatic Cobb Measurement of Scoliosis Based on Fuzzy Hough Transform with Vertebral Shape Prior

Abstract

To reduce variability of Cobb angle measurement for scoliosis assessment, a computerized method was developed. This method automatically measured the Cobb angle on spinal posteroanterior radiographs after the brightness and the contrast of the image were adjusted, and the top and bottom of the vertebrae were selected. The automated process started with the edge detection of the vertebra by Canny edge detector. After that, the fuzzy Hough transform was used to find line structures in the vertebral edge images. The lines that fitted to the endplates of vertebrae were identified by selecting peaks in Hough space under the vertebral shape constraints. The Cobb angle was then calculated according to the directions of these lines. A total of 76 radiographs were respectively analyzed by an experienced surgeon using the manual measurement method and by two examiners using the proposed method twice. Intraclass correlation coefficients (ICC) showed high agreement between automatic and manual measurements (ICCs > 0.95). The mean absolute differences between automatic and manual measurements were less than 5°. In the interobserver analyses, ICCs were higher than 0.95, and mean absolute differences were less than 5°. In the intraobserver analyses, ICCs were 0.985 and 0.978, respectively, for each examiner, and mean absolute differences were less than 3°. These results demonstrated the validity and reliability of the proposed method.