Table 2 Performance of model trained on sites 0 and 1 versus a model trained on all sites.

From: Deep COVID DeteCT: an international experience on COVID-19 lung detection and prognosis using chest CT

Test Site

Institution

Train/Val. Sites

Normals AUC

COVID− PNA AUC

COVID+ AUC

3

Stanford

0,1

0.898

0.741

0.906

3

Stanford

0,..,2,4,..,11

0.952

0.831

0.93

4

Unity Health

0,1

0.844

0.603

0.859

4

Unity Health

0,..,3,5,..,11

0.98

0.829

0.914

5

Koç

0,1

0.875

0.351

0.822

5

Koç

0,..,4,6,..,11

0.948

0.776

0.909

6

Rajaie

0,1

0.903

0.775

0.724

6

Rajaie

0,..,5,7,..,11

0.984

0.811

0.858

8

UNIFESP

0,1

0.929

0.347

0.928

8

UNIFESP

0,..,7,9,..,11

0.987

0.895

0.916

9

Henry Ford

0,1

0.967

0.620

0.874

9

Henry Ford

0,..,8,10,..,11

0.986

0.830

0.889

10

TUMS-1

0,1

0.948

N/A

0.947

10

TUMS-1

0,..,9,11

0.978

N/A

0.933

11

MosMedData

0,1

0.636

N/A

0.604

11

MosMedData

0,..,10

0.806

N/A

0.808

Back to article page