Reliability of 3D reconstruction of the spine of mild scoliotic patients

Study Design. A reliability study was conducted in quantitative 3-dimensional (3D) measurements for mild scoliosis. Objective. To evaluate the intrarater and interrater reliability of a computer tool used for 3D reconstruction of the spine. Summary of Background Data. No reliability study of spinal in vivo 3D medical imaging measurements has been performed in the literature. Methods. This study included 30 patients (mean age 13 years) with mild idiopathic scoliosis. Spinal 3D reconstruction was performed using a new technique called semiautomatic 3D reconstruction, which requires only the location of the corners of each vertebral body on 2 orthogonal views. Three raters performed the 3D reconstruction procedure on the 30 pairs of radiographs in random order. One of the raters repeated the procedure for the 30 patients 15 days later. Inter-reliability and intra-reliability were estimated for different parameters: thoracic kyphosis, lumbar lordosis, Cobb's angle, pelvic morphologic and positional parameters, and axial rotation. Results. Intraclass correlation coefficient showed good or very good agreement for most of the measurements. The 95% prediction limits are approximately 4 for the measurements of spinal curves, 2 for pelvic parameters, and axial vertebral rotation. Conclusions. The reliability of 3D reconstruction of the spine is acceptable, and this technique can be used for clinical studies.