作者: Christian Bauer , Melissa A. Krueger , Wayne J. Lamm , Brian J. Smith , Robb W. Glenny
DOI: 10.1109/TBME.2013.2277936
关键词: Sensitivity (control systems) 、 Lumen (anatomy) 、 Computer vision 、 Airway tree 、 Segmentation 、 Image segmentation 、 Point (geometry) 、 Path (graph theory) 、 Computer science 、 Artificial intelligence
摘要: A highly automated method for the segmentation of airways in serial block-face cryomicrotome images rat lungs is presented. First, a point inside trachea manually specified. Then, set candidate airway centerline points automatically identified. By utilizing novel path extraction method, between root tree and each obtained. Local disturbances are robustly handled by approach, which avoids shortcut problem standard minimum cost algorithms. The union all paths utilized to generate an initial structure, pruning algorithm applied remove erroneous subtrees or branches. Finally, surface used obtain lumen. was validated on five image volumes Sprague-Dawley rats. Based expert-generated independent standard, assessment identification lumen performance conducted. average detection sensitivity 87.4% with 95% confidence interval (CI) (84.9, 88.6)%. plot as function radius provided. combined estimate specificity 100% CI (99.4, 100)%. number diameter terminal branches 1179 159 μm, respectively. Segmentation results include up 31 generations. regression intercept slope measurements derived from final segmentations were estimated be 7.22 μm 1.005, developed approach enables quantitative studies physiology lung diseases rats, requiring detailed geometric models.