Statistical Shape Model Construction of Lumbar Vertebrae and Intervertebral Discs in Segmentation for Discectomy Surgery Simulation

Discectomy procedure simulations require patient-specific and robust three-dimensional representation of vertebral and intervertebral disc structures, as well as existing pathology, of the lumbar spine. Prior knowledge, such as expected shape and variation within a sample population, can be incorporated through statistical shape models to optimize the image segmentation process. This paper describes a framework for construction of statistical shape models (SSMs) of nine L1 vertebrae and eight L1-L2 intervertebral discs from computed tomography and magnetic resonance (MR) images respectively. The generated SSMs are utilized as a reference for knowledge-based priors to optimize coarse-to-fine multi-surface segmentation of vertebrae and intervertebral discs in volumetric MR images. Correspondence between instances within each model has been established using entropy-based energy minimization of particles on the image surfaces, which is independent of any reference bias or surface parameterization techniques. The resulting shape models faithfully capture variability within the first seven principal modes.