A Statistical Shape Model Pipeline to Enable the Creation of Synthetic 3D Liver Data

Abstract The application of machine learning approaches in medical technology is gaining more and attention. Due to the high restrictions for collecting intraoperative patient data, synthetic data increasingly used support training artificial neural networks. We present a pipeline create statistical shape model (SSM) using 28 segmented clinical liver CT scans. Our consists four steps: preprocessing, rigid alignment, template morphing, modeling. compared two different morphing approaches: Laplace-Beltrami-regularized projection (LBRP) nonrigid iterative closest points translational (N-ICP-T) evaluated both their corresponding performance six metrics. LBRP achieved smaller mean vertex-to-nearest-neighbor distances (2.486±0.897 mm) than N-ICP-T (5.559±2.413 mm). Generalization specificity errors were consistently lower those N-ICP-T. first principal components SSM showed realistic anatomical variations. was comparable state-of-the-art model.