Table 3 Agreement and accuracy of model versus pathologists grading, overall as well as split by nuclear pleomorphism, tubular differentiation and mitotic count.

From: Deep learning-based breast cancer grading and survival analysis on whole-slide histopathology images

Target

Cohen’s Kappa (SD)

Accuracy (SD)

Tumor grade

0.59 (± 0.04)

0.80 (± 0.02)

Nuclear score

0.41 (± 0.04)

0.70 (± 0.02)

Tubular score

0.35 (± 0.04)

0.70 (± 0.02)

Mitoses score

0.53 (± 0.04)

0.77 (± 0.02)

  1. Results are shown on the test set (n = 686). The standard deviation (SD) was calculated using bootstrapping.
Back to article page