Fig. 1.

Study design of the CNN model development and application. (a) A CNN model (VGG-19) was pre-trained on ImageNet dataset, and transfer learning was used to train the CNN model with the training set. Two independent image data sets were used to assess the classification accuracy of the model. (b) HE WSI image (20 × magnification) was segmented by the CNN model with sliding window methods. Eight tissue classes (excluding BAC class) ratio was calculated by counting each tissue area in the segmented result, and the tumour-stroma ratio was also obtained. CNN, convolutional neural network; HE, haematoxylin–eosin; WSI, whole-slide image; ADI, adipose; BAC, background; DEB, debris; LYM, lymphocyte aggregates; MUC, mucus; MUS, muscle; NOR, normal mucosa; STR, stroma; TUM, tumour epitheliu.