Fig. 1: Method overview for automatic detection of Barrett’s esophagus (BE) from H&E and TFF3 slides, including dataset preprocessing, model training, and data used in the study. | Nature Communications

Fig. 1: Method overview for automatic detection of Barrett’s esophagus (BE) from H&E and TFF3 slides, including dataset preprocessing, model training, and data used in the study.

From: Enabling large-scale screening of Barrett’s esophagus using weakly supervised deep learning in histopathology

Fig. 1

a H&E and TFF3 stained histopathological slides are scanned from adjacent sections and preprocessed using distinct pipelines for H&E and TFF3, as shown in purple and blue boxes, respectively. b Preprocessed slides are split on-the-fly into non-overlapping tiles and used to train weakly supervised BE-TransMIL models end-to-end from H&E and TFF3 slides separately. A similar training procedure is performed for both stains.

Back to article page