Table 1 Number of WSIs present in the individual (2nd–4th columns), combined (5th-6th columns) and independent test (7th column) datasets where \({D}_{{cleo}}\), \({D}_{{fin}},{D}_{{tcga}}\), and \({D}_{{tiger}}\) correspond to the CLEOPATRA28, FinHER29, The Cancer Genome Atlas Breast Invasive Carcinoma cohort (TCGA-BRCA) and TIGER datasets while \({D}_{{mix\_FFPE}}\) and \({D}_{{mix\_all}}\) correspond to the combined datasets

From: Annotation-efficient deep learning for breast cancer whole-slide image classification using tumour infiltrating lymphocytes and slide-level labels

Dataset

\({D}_{{cleo}}\)

\({D}_{{fin}}\)

\({D}_{{tcga}}\)

\({D}_{{mix\_FFPE}}({D}_{{cleo}}+{D}_{{fin}})\)

\({D}_{{mix\_all}}({D}_{{cleo}}+{D}_{{fin}}+{D}_{{tcga}})\)

\({D}_{{tiger}}\)(Test only)

Tissue Type

FFPE

FFPE

FF

FFPE

FFPE + FF

FFPE

Train

477

416

444

893

1337

Validation

120

104

112

224

336

Test

150

131

139

281

420

82