Assessment of scoliosis using 3-D ultrasound volume projection imaging with automatic spine curvature detection

Abstract

X-ray imaging is a gold standard to diagnose scoliosis, a medical condition defined as lateral spine curvature > 10°. However, radiation hazard restricts its application. 3-D ultrasound imaging shows potential for radiation-free scoliosis assessment. Recently, ultrasound volume projection imaging (VPI) was reported to provide a coronal view of spine for manual measurement of spine curvature. An automatic method for processing VPI images is now very much desired to avoid using tedious and subjective manual procedures, especially for scoliosis mass screening. In this paper, we reported an automatic method to quantitatively determine the spine curvature using the spine column profile, represented by the darkest region along the midline of VPI image. The method started with image enhancement using histogram equalization, and identified the points by locating the darkest points row by row in the enhanced image, followed by a polynomial curve fitting to the detected points. The spine curvature angle was finally calculated according to the inflection points on the curve. The performance was tested on the VPI images obtained from 36 subjects (Age: 30.1 ± 14.5) with different spine curvatures. The curvatures obtained using the automatic method had a significant correlation with those by the manual method (r = 0.92; p <; 0.001). The proposed automatic method is capable of measuring the spine curvature on the VPI images, thus greatly facilitating the use of 3-D ultrasound for scoliosis assessment.