⚠️ Educational illustration only — not a diagnostic tool. Real pN staging requires a pathologist's review of an actual whole-slide specimen.
This simulates how a whole-slide-image AI model, of the kind benchmarked on the CAMELYON lymph-node metastasis detection challenge (Bejnordi et al.), scans a grid of tissue patches from a lymph-node section and flags which patches look tumor-positive. The simulated 10×10 patch grid represents a small field of a lymph-node whole-slide image; the total metastatic extent is then classified into the real AJCC pN staging bands used in breast-cancer pathology reports.
Each patch on the grid represents roughly a 0.2 mm × 0.2 mm tile of tissue, so the full 10×10 grid stands for a simulated 2.0 mm × 2.0 mm field of view. Move the tumor-positive fraction slider to set how much of that field is flagged tumor-positive, and switch the pattern between "scattered" (isolated positive patches, illustrating isolated tumor cells) and "clustered" (a single contiguous growing deposit, illustrating a micro- or macrometastasis). The simulation finds the largest contiguous group of positive patches and measures its extent against the real AJCC size thresholds: isolated tumor cells (ITC) at or below 0.2 mm, micrometastasis from 0.2 mm to 2.0 mm, and macrometastasis above 2.0 mm.
Tumor-positive patch fraction slider, scattered/clustered spatial pattern selector, and a resample button to regenerate a new random layout at the same settings.
Did you know that the CAMELYON16 challenge found that the best-performing AI algorithms could match or approach the accuracy of expert pathologists at detecting lymph-node metastases in whole-slide images, and that combining an algorithm's output with a pathologist's review outperformed either working alone? A single lymph node cross-section can span thousands of high-power microscope fields, which is exactly the exhaustive-search problem this kind of AI-assisted screening is designed to help with.
This simulates how a whole-slide-image AI model, of the kind benchmarked on the CAMELYON lymph-node metastasis detection challenge (Bejnordi et al.), scans a grid of tissue patches from a lymph-node section and flags which patches look tumor-positive, then classifies the largest contiguous deposit into the real AJCC pN staging bands used in breast-cancer pathology reports.
This simulates how a whole-slide-image AI model, of the kind benchmarked on the CAMELYON lymph-node metastasis detection challenge (Bejnordi et al.), scans a grid of tissue patches from a lymph-node section and flags which patches look tumor-positive, then classifies the largest contiguous deposit into the real AJCC pN staging bands used in breast-cancer pathology reports.
Move the tumor-positive fraction slider to set how much of the simulated field is flagged tumor-positive, and switch the pattern between "scattered" (isolated positive patches, illustrating isolated tumor cells) and "clustered" (a single contiguous growing deposit, illustrating a micro- or macrometastasis). Use the resample button to regenerate a fresh random layout at the same settings, and watch the pN category badge update live against the real AJCC size thresholds.
Did you know that the CAMELYON16 challenge found that the best-performing AI algorithms could match or approach the accuracy of expert pathologists at detecting lymph-node metastases in whole-slide images, and that combining an algorithm's output with a pathologist's review outperformed either working alone? A single lymph node cross-section can span thousands of high-power microscope fields, which is exactly the exhaustive-search problem this kind of AI-assisted screening is designed to help with.