Biomechanical deformable image registration of longitudinal lung CT images using vessel information

摘要: Spatial correlation of lung tissue across longitudinal images, as the patient responds to treatment, is a critical step in adaptive radiotherapy. The goal this work expand biomechanical model-based deformable registration algorithm (Morfeus) achieve accurate presence significant anatomical changes. Six cancer patients previously treated with conventionally fractionated radiotherapy were retrospectively evaluated. Exhale CT scans obtained at treatment planning and following three weeks treatment. For each patient, was registered follow-up using Morfeus, algorithm. To model complex response lung, an extension Morfeus has been developed: initial deformation estimated consisting boundary conditions on chest wall incorporating sliding interface lungs. It hypothesized that addition based vessel tree matching would provide robust reduction residual error. this, trees segmented two images by thresholding vesselness image Hessian matrix's eigenvalues. point reference centerline, displacement vector applying variant Demons between deformed CT. An expert independently identified corresponding landmarks well distributed compute target errors (TRE). TRE was: [Formula: see text], text] mm after rigid registration, tree, respectively. In conclusion, vessels significantly improved accuracy modeling tumor over course Minimizing these geometrical uncertainties will enable future plan adaptation strategies.