Database Annotation with few Examples: An Atlas-based Framework using Diffeomorphic Registration of 3D Trees

Automatic annotation of anatomical structures can help simplify workflow\nduring interventions in numerous clinical applications but usually involves a\nlarge amount of annotated data. The complexity of the labeling task, together\nwith the lack of representative data, slows down the development of robust\nsolutions. In this paper, we propose a solution requiring very few annotated\ncases to label 3D pelvic arterial trees of patients with benign prostatic\nhyperplasia. We take advantage of Large Deformation Diffeomorphic Metric\nMapping (LDDMM) to perform registration based on meaningful deformations from\nwhich we build an atlas. Branch pairing is then computed from the atlas to new\ncases using optimal transport to ensure one-to-one correspondence during the\nlabeling process. To tackle topological variations in the tree, which usually\ndegrades the performance of atlas-based techniques, we propose a simple\nbottom-up label assignment adapted to the pelvic anatomy. The proposed method\nachieves 97.6\\% labeling precision with only 5 cases for training, while in\ncomparison learning-based methods only reach 82.2\\% on such small training\nsets.\n