Prediction of spinal deformity progression with serial spine images and advanced numerical modeling techniques

Adolescent idiopathic scoliosis (AIS) is a three-dimensional (3D) spine and trunk deformity associated with pain, osteoarthritis, and in rare cases death via cardiopulmonary complications. Previous studies have indicated that various factors have been associated with curve progression. When the causes of progression of spinal deformities were investigated, comparisons of prognostic factors such as age, curve magnitude, and growth velocity were commonly assessed on the changes in data between the first visit and last follow-up. Therefore, the use of inconsistent time-intervals of historic data and progression thresholds of Cobb angle could alter the indication of progression. In addition, these previous studies were limited to simply indicating the possibility of progression of a scoliotic curve. It is still not clear to what extent these factors can be used for the prediction of the future spinal deformity. The studies in this thesis detected potential prognostic factors by using time-series data sets of at least two six month follow-up intervals, based on both progression thresholds of Cobb angles of 5° and 10°. The AIS progression patterns in time were extracted, and their effectiveness in assisting the prediction of spinal deformities was investigated. The predictions of future spinal deformities in the 2D spinal indices of Cobb angle and apical lateral deviation were made using advanced numerical modeling techniques, including generalized cross-validation (GCV), artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS). The artificial progression surface (APS) technique was used to predict the future deformed 3D spine. The present studies may assist orthopaedic surgeons with decisions regarding future scoliosis treatments. The present methodologies for the prediction of spinal deformities with serial spinal images can be used as a foundation for techniques to predict future torso deformities with serial deformed torso images, as well as the future spinal deformities with serial deformed torso images.