Fig. 1 | Scientific Data

Fig. 1

From: Inter-species cell detection - datasets on pulmonary hemosiderophages in equine, human and feline specimens

Fig. 1

Overview of the macrophage annotation and validation pipeline: The publicly available RetinaNet object-detection model trained on equine slides4 is used to perform inference on the unannotated slides, followed by a semi-automatic clustering step which clusters cells by size. Error-prone cells are highlighted and can then be efficiently deleted by a human expert. Afterwards, a human expert screens all WSI to increase the dataset consistency. Finally, a regression-based clustering system is applied to support experts searching for misclassifications of the hemosiderin grade.

Back to article page