Table 7 Overview of SKET performance, reporting the precision, the recall and the F1-score of the single classes, in both Catania and Radboudumc reports.

From: Unleashing the potential of digital pathology data by training computer-aided diagnosis models without human annotations

Class

Precision

Recall

F1-score

Support

SKET performance per class (Catania)

  Cancer

0.84

0.94

0.89

379

  High-Grade Dysplasia

0.90

0.92

0.91

454

  Low-Grade Dysplasia

0.75

0.90

0.82

529

  Hyperplastic polyp

0.68

0.94

0.79

181

  Normal glands

0.87

0.92

0.89

438

SKET performance per class (Radboudumc)

  Cancer

0.94

0.95

0.94

188

  High-Grade Dysplasia

0.66

0.83

0.73

94

  Low-Grade Dysplasia

0.85

0.81

0.83

453

  Hyperplastic polyp

0.84

0.97

0.90

428

  Normal glands

0.92

0.88

0.90

1048

  1. The support of the class represents the number of true positive cases for a particular class.