Skull Reconstruction Using Statistical Shape Models

Abstract: In case of a head injury or trauma a surgeon opens up the skull by removing a bone piece of the skull to either relieve pressure or get access to brain tissue. The first option is to follow up with putting back the same skull piece. In the circumstances this is not possible however, a surgeon must open up the skull and form a piece using bone cement that fits the whole during the surgery, which is time consuming This leads to an increase risk of infection and other medical risks for the patient. This project investigates the possibility to mathematically reconstruct the piece missing in the skull, giving the surgeon the option to already have a fitted piece ready before the operation begins. The goal of the project was to investigate if a statistical shape model could be used for cranial reconstruction. The model was built from a data set from Skåane University Hospital, it was preprocessed using Medviso Segment 3D Print software and all other implementation was done in Matlab. Each skull were sampled to a point cloud, with skull radii and thickness as parameters. The shape model was built from a point cloud sampled from each skull. The skulls were registered to each other using ICP registration. The resulting model reconstruct successfully a damaged skull with 1 mm average error. Six modes were enough to account for 90% of the shape variability. The results of average error of 1 mm was not deemed enough to be clinically useful.