Antibody Supervised Training of a Deep Learning Based Algorithm for Leukocyte Segmentation in Papillary Thyroid Carcinoma

The quantity of leukocytes in papillary thyroid carcinoma (PTC) potentially have prognostic and treatment predictive value. Here, we propose a novel method for training convolutional neural network (CNN) algorithm segmenting PTCs. Tissue samples from two retrospective PTC cohort were obtained representative tissue slides twelve patients stained with hematoxylin eosin (HE) digitized. Then, the HE destained restained immunohistochemically (IHC) antibodies to pan-leukocyte anti CD45 antigen scanned again. stain-pairs all registered, image tiles regions interests exported. processed 3,3'-diaminobenzidine (DAB) areas representing expression turned into binary masks. These masks applied as annotations on used CNN algorithm. Ten whole slide images (WSIs) using five-fold cross-validation remaining an independent test set trained model. For visual evaluation, was run WSIs, total 238,144 sized 500 × pixels analyzed. had intersection over union 0.82 detection when comparing prediction ground truth mask. We conclude that this generating antibody supervised destain-restain IHC guided resulted high accuracy segmentations images.