Measurement of Vertebral Rotation from Moire Image for Screening of Adolescent Idiopathic Scoliosis

Abstract

Periodic spine screening in teenagers is important for early detection of adolescent idiopathic scoliosis, a deformity characterized by an abnormal spinal curve and vertebral rotation. Conventional methods for measuring vertebral rotation involve exposure to radiation and can be complicated by challenges inherent to manual identification of spinal features. We propose a method for automatic measurement of vertebral rotation using a nonradiation method that is easy to administer and is not associated with risk. To measure the angle of vertebral rotation from a Moire image, which is nonradiation method, convolutional neural network is applied to estimate the spinal position. In addition, to measure the angle from the spinal position in the transverse plane, we generate a spinal model of a teenager to obtain spinal dimensions in the transverse plane. The angle of vertebral rotation method is valid, having high correlation with method using computed tomography. The angle of vertebral rotation is meaningful for pathological diagnosis of adolescent idiopathic scoliosis.