Előadás címe: From Bone Segmentation to Customized Implants
Helyszín: Online
Időpont: 2020.06.30., 14:00 – 14:20
Kivonat: For the extraction of bone structures from computer tomography images, an ensemble of segmentation algorithms with optimal parameter setting is applied, whose outputs are aggregated by some decision rule (e.g. majority voting). To reduce the computational costs and time, nearest neighbor image pyramid-based noisy energy function evaluation method is proposed for the local search technique simulated annealing (at different downscaling levels during the process). To ensure the convergence of SA, we have to apply a strategy to select the appropriate downscaling level (i.e. the resolution of the images) in each iteration to control the standard deviation of the noise according to the actual temperature. The reconstruction of skull defect can be based on e.g. cubic Bezier curves, discretization of curvature, principal curvatures and torsion, or deep learning methods.