Direct Cellularity Estimation on Breast Cancer Histopathology Images Using Transfer Learning

Figure 5

Illustration of our KNN method to map the ranking scores to cellularity. The red dots represent the training samples that lie within the 30-nearest neighborhood of a testing samples with ranking score 4.0. The cyan dots represent those outside the neighborhood. The prediction for the testing sample is the mean of cellularity of all the red dotted samples, as denoted by the green star.