Table 1 Description of evaluation datasets

From: ROSIE: AI generation of multiplex immunofluorescence staining from histopathology images

Evaluation Datasets

In training data?

# Samples

# Slides

# Cells

Study

Disease

Stanford-PGC

Yes

Yes

149

1

817,765

Ochsner-CRC

No

No

94

1

635,649

Tuebingen-GEJ

No

No

240

1

365,734

UChicago-DLBCL

Yes

Yes

2

1

3,099,419

Total

-

-

485

8

4,918,567

  1. We describe the four evaluation datasets used in this study. Samples from Stanford-PGC and UChicago-DLBCL are divided into training and test splits; on the other hand, no samples and disease types from Ochsner-CRC and Tuebingen-GEJ are used in the model’s training data. UChicago-DLBCL contains two full tissue samples, whereas the rest of the datasets consist of TMA core samples.
Back to article page