Ultra-short echo-time magnetic resonance imaging lung segmentation with under-Annotations and domain shift

Fumin Guo, Dante P. I. Capaldi, David G. McCormack, Aaron Fenster, Grace Parraga

Abstract: Highlights•K-means for pairwise high-dimensional feature clustering in kernel space with spatially continuous regularization for lung segmentation in UTE MRI.•CNNs trained on small datasets with under-annotations and tested on UTE MRI datasets from two centres.•Atlas-based segmentation using three atlas images without requiring complex atlas generation, label fusion, or heavy computational burden.•Excellent UTE lung segmentation accuracy, precision, and generalizability that outperformed some state-of-the-art CNNs and similar to repeated manual segmentation.

External IDs:dblp:journals/mia/GuoCMFP21