Design and evaluation of decision support system to predict spinal deformity in idiopathic scoliosis, using image processing techniques

Abstract

Abstract Introduction: Scoliosis is known as the most common spine disorder and is defined as a three-dimensional deformity of the vertebral column, which is accompanied by vertebral curvature and rotation. The measure of the value of the curvature of the vertebrae is calculated using radiographs and by an index called Cobb’s Angle, and also the measure of the degree of the vertebral rotation is calculated by specialists and using CT scan images. Objectives: The goal of this study is to computerize the process of measuring the curvature and rotation of the vertebrae using a radiographic image, in order for it to ultimately lead to increased accuracy and reduced time of measurement. Methods and Materials: Using radiographic images and preprocessing and image processing techniques including the use of Canny edge detector, noise elimination, Hough transform, and RSF technique the first and the last vertebrae of the curvature of the scoliosis were extracted and their angles were calculated automatically and semi-automatically. After automatic detection of apical vertebra, the standard Nash Moe method was used to calculate the degree of the rotation of the vertebra. Ninety-eight radiographic images were evaluated by five specialists and within two weeks, and the repeatability and reproductibility were evaluated and compared using MAD (mean absolute difference), SEM (standard error of measurement), and ICC (intra-class correlation coefficient) statistical indices for the curvature and rotation of the vertebrae in both manual and computerized measurements. And the Bland-Altman plot was used to assess the degree of the agreement between the evaluators. Also, the mean measurement times of computerized and manual methods were compared using the paired t-test. Findings: The results obtained from the assessment of the repeatability in the computerized measurement of Cobb’s angle for the MAD, SEM, and ICC statistical indices were calculated equal to 1.42, 2.11 and 0.97, and 1.83, 1.43 and 0.97, for the automatic and the semi-automatic methods, respectively. Also, for the measurement of the degree of the vertebra rotation, the above indices were equal to 1.84 and 1.06 and 0.95, respectively. The results obtained from the assessment of the reproductibility in the computerized measurement of Cobb’s angle for the MAD, SEM, and ICC statistical indices were calculated equal to 2.12, 1.24 and 0.97, and 2.22, 2.09 and 0.95, for the automatic and the semi-automatic methods, respectively. In the calculation of the reproductibility of the vertebral rotation, these indices were 2.04 and 1.46 and 0.96, respectively. The results of the Bland-Altman plot showed a 95% agreement between the observers. The results of the paired t-test indicated a significant reduction in the mean computerized measurement time of the Cobb's angel and vertebral rotation compared to the manual method . Conclusion: Computer-based methods for the evaluation of scoliosis are more accurate than the manual measurement method and are associated with reduced measurement times. Keywords: Ideiopatic Scoliosis, Cobb angle, Vertebrea rotation, Repeatability, Reproducyibility