We propose an automatic nodule registration method between baseline and follow-up chest CT scans. Initial alignment using the center of the lung volume corrects the gross translational mismatch, and rigid registration using coronal and sagittal maximum intensity projection images effectively refines the rigid motion of the lungs. Nodule correspondences are established by finding the most similar region in terms of density as well as the geometrical constraint. The proposed nodule registration method increased the nodule hit rate (the ratio of the number of successfully matched nodules to total nodule number) from 26% to 100%.
Keywords: Computed tomography; Follow-up CT study; Geometrical constraint using log-polar image; Nodule template matching; Pulmonary nodule registration; Rib matching.
© 2013 Published by Elsevier Ltd.