Table 6 The confusion matrix on patient level Hematoxylin and Eosin stained images.

From: Quantitative diagnosis of breast tumors by morphometric classification of microenvironmental myoepithelial cells using a machine learning approach

Prediction

Actual

Normal

UDH

LG-DCIS

HG-DCIS

Normal

5

1

0

1

UDH

0

4

1

0

LG-DCIS

0

0

5

0

HG-DCIS

0

0

1

4

  1. 4 groups: Accuracy 81.8% (kappa index 0.76). Normal/Benign vs DCIS: Accuracy 90.9% (kappa index 0.82).