Skip to main content
Fig. 2 | Biological Procedures Online

Fig. 2

From: Development of an Interpretable Deep Learning Model for Pathological Tumor Response Assessment After Neoadjuvant Therapy

Fig. 2

Color differences of WSIs from 2 cohorts. a Representative WSIs from Cohort 1 (Pudong-Branch 2020–2021) and External Cohort 2 (Puxi-Branch 2022) b Bin counts of red, green, blue channel values within tumor bed regions, highlighting quantitative variation between 2 cohorts. c Patch-level dataset generation workflow. Annotated tumor beds from pathologists were extracted from WSIs at 20 × magnification. Non-overlapping 256 × 256 pixel patches were exhaustively cropped from the tumor bed regions. The number of patches extracted per WSI ranged from 7 for small tumor beds to 21,737 for large tumor beds, depending on the annotated tumor bed size within each whole slide image

Back to article page