Funding outcome

Adolescent Idiopathic Scoliosis (AIS) is a three-dimensional (3D) deformity of the spine that occurs in 2-3% of adolescents. Cobb angle, vertebral rotation and kyphotic angle which are 2D measurements from frontal and sagittal planes' radiographs, are the most common methods used to describe scoliosis severity, to monitor the progression of a curve, to decide on a treatment method, and to evaluate treatment outcomes. From the literature, approximately 30% of patients with AIS require active treatment which includes bracing and surgery, while 70% of patients are only monitored until skeletal maturity. Although orthopedic surgeons know radiation exposure is not desirable for the growing child, there is no other commonly used non-ionizing imaging method for scoliosis. My team developed a non-radiation imaging method, ultrasound, which can image the spine and allow us to measure the spinal curvature accurately and reliably on the ultrasound images. However, human bias and measurement errors exist regardless of the imaging methods. Especially, the measurements from ultrasound images require longer training and practice. Furthermore, to obtain 3D spinal deformity information from radiography, the frontal and lateral radiographs should be acquired simultaneously. Acquiring lateral radiographs expose children to higher ionizing radiation than acquiring the posteroanterior radiographs. The advantage of using ultrasonography over radiography is a single ultrasound scan can provide 3D curvature information. As applying artificial intelligence on medical images for disease diagnosis becomes promising, it is proposed to apply this for scoliosis application. This project aims to apply artificial intelligence (Al) to process the radiographs and ultrasound images so that fully automated spinal curvature measurements can be obtained to minimize human measurement bias and errors.