Assessment of 3D reconstruction of scoliotic human torso using optical imaging techniques

A spine affected by scoliosis shows evidence of a lateral or sideways curvature and rotation of the vertebrae. Once diagnosed, the clinical assessment of scoliotic patients is completed with entire spinal X-rays every six months, leading to deleterious biological effects. A non-invasive technique was developed to detect spinal curve deformities from precise three-dimensional (3D) computerized torso surface asymmetry measurements, quantified initially with a laser imaging system and a 3D stereoradiographic reconstruction using Artificial Neural Networks (ANN). To enable integration of more accurate data (∼ 1mm) with faster data acquisition, a new imaging system (Inspeck, Montreal) was implemented. The purpose of this research was to adapt the patient positioning devices to be compatible with the combined imaging system modalities, as well as to evaluate the overall system accuracy and assess the impact of the new torso imaging system on the quantification of torso surface indices. The implementation of the Inspeck system clinically required a detailed investigation of the imaging system operation, development of an experimental design to evaluate factors that influence the 3D torso reconstruction accuracy, and determination of the optimal set up in laboratory and hospital conditions. The 3D reconstructed models, were evaluated technically (accuracy, repeatability, reliability), and subsequently implemented in algorithms developed to determine for the torso indices. Study results provide a strong foundation for enabling integration of the torso indices obtained with the Inspeck technology into both the ANN model, and routine clinical assessments of changes in torso shape for diagnosis of scoliosis, as well as monitoring and prediction of scoliosis progression.