Fig. 1: Machine learning model annotation collection, training, and application. | npj Precision Oncology

Fig. 1: Machine learning model annotation collection, training, and application.

From: AI powered quantification of nuclear morphology in cancers enables prediction of genome instability and prognosis

Fig. 1

a Model workflow. Briefly, pathologists trained expert annotators to perform exhaustive annotations of nuclei on H&E slide patches from diverse tissue sources. These were used to train a pan-H&E nucleus detection and segmentation model, which was subsequently evaluated on held-out patches and applied to exhaustively segment nuclei in three WSI datasets. b Features extracted from the model. Mean and standard deviation values were calculated for these features at the whole-slide level for cancer cells, lymphocytes, and fibroblasts.

Back to article page