Whole slide histologic grading of breast cancer using convolutional neural networks.

e13607 Background: The Nottingham histologic grade (NHG) is a strong prognostic factor in early-stage breast cancer. It consists of nuclear pleomorphism, tubular formation, and mitotic count. We recently developed an artificial intelligence (AI) based automatic grading system. Methods: In this study, we have retrospectively evaluated 179,651 hematoxylin eosin-stained patches extracted from 402 digitized biopsies 338 patients with confirmed invasive ductal carcinoma diagnosis. data was collected at Acıbadem University Hospital between 2017 2021. slides were manually labeled by seven pathologists before being used to train the deep learning models (DL). pre-trained (on ImageNet) DL architectures which are EfficientNet backbone U-Net, YOLOv5, DenseNet161, modified VGG-11 been fine-tuned study’s dataset for tubule segmentation, nuclei detection, mitosis classification, pleomorphism classification tasks, respectively. Data augmentation boosting accuracy done. Semantic segmentation object detection nuclei, image count also performed. Results: AI-based algorithms obtained reproducible scores mean F1 scores, sensitivities, specificities as presented Table. Conclusions: system accurate evaluation components NHG. This will speed up pathology workflow clinic, provide decision support mitigate sensitivity associated traditional process. [Table: see text]