Sparse Modeling of Mandibular Reconstruction Procedures Using Statistical Geometric Features

This paper introduces a sparse modeling method that uses statistical geometric features for automated pre-operative planning. It further shows the application of this method to mandibular reconstruction with fibular segments. With this method, instead of using all the training data, only a small number of data that have similar features to the test data are selected and appropriately synthesized to reconstruct patient-specific plans. We compared the performance of three automated planning models using 120 patterns of mandibular reconstruction data manually planned by oral surgeons. The sparseness of the data selection and the efficacy of the automated planning framework were quantitatively confirmed.