Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Ontology-guided segmentation and object identification for developmental mouse lung immunofluorescent images

Fig. 2

ROC curves for labels generated by XGBoost as compared with ground truth labels assigned by a histopathologist. In the description below the term “region” refers to segmentations generated by the odifmap pipeline. The probe set in the selected images targeted the distal acinar tubule bud. The three curves show results based on different acceptance criteria for whether a region correctly identifies the structure. a Only a single region with > 50% overlap of a whole structure as labeled by the pathologist is accepted as correct. All other regions, regardless of label, are marked incorrect. b When more than one region overlaps the histopathologist-labeled structure, these regions are manually merged, and the single merged region is accepted as correct. c When more than one region overlaps the histopathologist-labeled structure, the individual regions are all accepted as correct

Back to article page