Automatic analysis method of 3D images in patients with scoliosis by quantifying asymmetry in transverse contours

Abstract

Scoliosis is characterized by a lateral curvature of spine in the coronal plane of at least 10°, showing three-dimensional deformities of the torso. Recently, a 3D body scanner was proposed as an ionizing radiation-free method, supplementary to clinical examinations and X-rays, for assessing scoliosis and its progression. Here we present an automatic method of analysis of the 3D images of the body torso delivered by the scanner, with the objective of capturing and characterizing deformities in the torso due to scoliosis. The method quantifies asymmetries in each of the torso contours, extracted from 2D cross sections of the 3D torso images, transverse to the vertical axis of the body. Three parameters were calculated: (1) circularity, (2) difference between the areas to left and right of the spinous process (LRAsm), and (3) difference between the ratios width/depth to the left and right of the centroid of the contour (ASR). The method was verified by analyzing twenty-six computed tomography images of patients with different types of scoliosis. In patients with thoracic scoliosis both LRAsm and ASR were larger, that is the asymmetry was stronger, in the thoracic than in the lumbar region, whilst in patients with lumbar scoliosis the inverse was obtained. Furthermore, larger values of LRAsm coincided with the apex position of scoliosis. The circularity factor did not capture scoliosis-related asymmetries. It may, however, be useful in localizing the vertebral level during analysis of 3D scanner data. The method can have the potential to use it for follow-up examinations of scoliosis, in addition to clinical examinations. Further validation of the method requires its application to 3D body scanner data, particularly for more cases of severe scoliosis, which have more irregular transverse contours and thus potential for improvement.