Breast Cancer Segmentation and Scoring in H&E


Logo for Breast Cancer Segmentation and Scoring in H&E

About

Image Version:
b069eac0-f3a5-4567-89ea-3b2013f9b109
Last updated:
Aug. 16, 2022, 6:49 p.m.
Inputs:
  • Generic Medical Image 
Outputs:
  • TIL Score  (Percentage of stromal area covered by tumour infiltrating lymphocytes. Values between 0 (percent) to 100 (percent).)
  • Breast Cancer Segmentation for TILs  (Class map with levels 1: invasive tumor, 2: tumor-associated stroma, 3: in-situ tumor, 4: healthy glands, 5: necrosis not in-situ, 6: inflamed stroma, 7: rest)
  • Lymphocyte Tumor Ratio  (Lymphocyte-Tumor-Ratio, a float between 0 and 1 quantifying the ratio of lymphocytes to tumor.)
  • Inflamed Tumor Ratio  (Inflamed Tumor Ratio, a float between 0 and 1 quantifying the proportion of tumor close to lymphocytes.)

Challenge Performance

Date Challenge Phase Rank
May 2, 2022 tiger Segmentation and Detection (Experimental) 287

Model Facts

Summary

Tumor infiltrating lymphocytes have been shown to have predictive value for survival and chemotherapy response prediction in breast cancer. We developed a fully automated deep-learning based algorithm to segment breast H&E tissue into the classes Tumor, Normal, Fat, Stroma, Necrosis and Lymphocytes and compute three lymphocyte based biomarkers including the (computational) tumor infiltrating lymphocytes score. Tumor infiltrating lymphocytes have been shown to have predictive value for survival and chemotherapy response prediction in breast cancer. We developed a fully automated deep-learning based algorithm to segment breast H&E tissue into the classes Tumor, Normal, Fat, Stroma, Necrosis and Lymphocytes and compute three lymphocyte based biomarkers including the (computational) tumor infiltrating lymphocytes score.

Citation: Aswolinskiy, Witali, et al. "Predicting pathological complete response to neoadjuvant chemotherapy in breast cancer from routine diagnostic histopathology biopsies." medRxiv (2022)

Mechanism

The algorithm has three steps:

  1. Preprocessing: Tissue is separated from background using a convolutional neural network
  2. Segmentation: The tissue is segmented into the classes Tumor, Normal, Fat, Stroma, Necrosis and Lymphocytes using a U-Net.
  3. From the segmentation, three biomarkers are computed: the computational tumor infiltrating lymphocytes score (TILs), inflamed tumor ratio (ITR) measuring the proportion of tumor within 80 microns of lymphocytes and the lymphocyte to tumor ratio (LTR).

Validation and Performance

Uses and Directions

This algorithm was developed for research purposes only. The terms of usage for Grand Challenge apply to the uploaded slides.

Warnings

Common Error Messages

Information on this algorithm has been provided by the Algorithm Editors, following the Model Facts labels guidelines from Sendak, M.P., Gao, M., Brajer, N. et al. Presenting machine learning model information to clinical end users with model facts labels. npj Digit. Med. 3, 41 (2020). 10.1038/s41746-020-0253-3