Model-based Sparse-to-dense Image Registration for Realtime Respiratory Motion Estimation in Image-guided Interventions
Respiratory motion is known to be an important problem in non-invasive image-guided tumor interventions that needs to be accounted for to achieve an accurate treatment delivery. To this end, we propose a novel motion estimation method that estimates dense motion fields in realtime for the entire treatment region based on intra-interventional image data acquired by state-of-the-art treatment systems like MRI-Linacs. Our method achieves state-of-the-art tracking accuracy (≈ 1 mm) at high frame rates by combining GPU-accelerated sparse feature point matching and patient-specific regularisation using a learned PCA-based motion model in a unified registration framework.
Evaluation of Six Registration Methods for the Human Abdomen on Clinically Acquired CT
This work evaluates current 3-D image registration tools on clinically acquired abdominal computed tomography (CT) scans. Thirteen abdominal organs were manually labeled on a set of 100 CT images, and the 100 labeled images were pairwise registered based on intensity information with six registration tools (FSL, ANTS-CC, ANTS-QUICK-MI, IRTK, NIFTYREG, and DEEDS). The results suggest that DEEDS yielded the best registration performance. All data and source code are available.