Fig. 1

Study workflow. A Top panel: A CNN model (CNN-HE) was used to classify the HE-stained WSI of colorectal cancer into eight tissue types and one slide background. A rough segmentation map was obtained. The Deep-TSR score is calculated as "the area of STR /the area of STR and TUM". Bottom panel: Using STR of tissue segmentation as the mask, we define the Deep-TIL score as the mean prediction probability of LYM class in STR class. B The Deep-immune score was synthesized by the Deep-TSR score and Deep-TIL score. Deep-TSR-high and Deep-TSR-low groups were given 0 and 1 points, respectively. Deep-TIL-low, Deep-TIL-middle, and Deep-TIL-high groups were given 1, 2, and 3 points, respectively. A four-level scoring system (score 1–4) was established by summing both the Deep-TSR score and the Deep-TIL score. HE, hematoxylin and eosin; WSI, whole-slide image; CNN, convolutional neural network; ADI, adipose; BAC, background; DEB, debris; LYM, lymphocyte aggregates; MUC, mucus; MUS, muscle; NOR, normal mucosa; STR, stroma; TUM, tumor epithelium; TSR, tumor-stroma ratio; TIL, tumor-infiltrating lymphocyte