Computer-assisted Cobb angle measurement from posteroanterior radiographs by a curve fitting method

Abstract

Incorrect determination of the top or end vertebrae of a scoliotic curve causes measurement errors when using the Cobb method. Such problems are due to subjectivity. This study was performed to minimize subjectivity by using a curve fitting method. A digitally scanned posteroanterior radiographic image was used for each patient. User-defined midpoints of the curvature, which indicate the midpoints of the vertebrae, were used to fit the polynomial equation. The locations of the top and end vertebrae were calculated using the first derivative of the fitted polynomial equation to obtain the analytical Cobb angle. The interobserver technical error of measurements was calculated as 1.06 degrees with a reliability coefficient of 1.00. The results indicate that 100% of the variations were not related to measurement error. Small angle variations show that subjectivity was minimized.