Fig. 1: Overview of annotation, deep learning system (DLS) development, and prognostic evaluation.
From: Deep learning models for histologic grading of breast cancer and association with disease prognosis
A Annotation Overview: Pathologists provided annotations at a region-level and slide-level for all components of the histologic grade, including identification of individual mitotic figures. B DLS overview: convolutional neural network models were developed for invasive carcinoma (INVCAR) as well as all three components of the histologic grading system. These patch-level models were used as input to stage 2 models to provide a component score at the slide level for each feature. C Prognostic evaluation: Component grade scores provided by the DLS or pathologists were used to fit Cox models for evaluation and comparison of prognostic value.
