Table 1 Target diseases and classifier performance for the gastroscopy dataset.

From: Weakly supervised end-to-end artificial intelligence in gastrointestinal endoscopy

ICD

Diagnosis

AUROC mean

AUROC 95% CI

N pos exams

N neg. exams

p-val

K29

Gastritis and duodenitis

0.698

0.007

9740

19,766

0.0000

K44

Diaphragmatic hernia

0.637

0.029

3725

25,781

0.0000

K21

Gastro-oesophageal reflux disease

0.619

0.023

3497

26,009

0.0000

K22

Other diseases of oesophagus

0.621

0.023

2619

26,887

0.0000

R10

Abdominal and pelvic pain

0.672

0.033

1419

28,087

0.0000

K25

Gastric ulcer

0.613

0.022

772

28,734

0.0000

CXX

Malignant neoplasm of oesophagus (C15) or stomach (C16)

0.761

0.048

763

28,743

0.0000

R13

Dysphagia

0.719

0.011

762

28,744

0.0000

K26

Duodenal ulcer

0.694

0.039

626

28,880

0.0000

C15

Malignant neoplasm of oesophagus

0.773

0.051

536

28,970

0.0000

B37

Candidiasis

0.703

0.126

469

29,037

0.0000

K31

Other diseases of stomach and duodenum

0.612

0.059

401

29,105

0.0001

I85

Oesophageal varices

0.650

0.003

332

29,174

0.0000

R12

Heartburn

0.587

0.107

312

29,194

0.0420

R11

Nausea and vomiting

0.648

0.068

279

29,227

0.0000

C16

Malignant neoplasm of stomach

0.693

0.066

245

29,261

0.0000

K92

Other diseases of digestive system

0.697

0.030

222

29,284

0.0000

D13

Benign neoplasm of other and ill-defined parts of digestive system

0.593

0.142

210

29,296

0.0948

D37

Neoplasm of uncertain or unknown behaviour of oral cavity and digestive organs

0.595

0.092

136

29,370

0.0728

D48

Neoplasm of uncertain or unknown behaviour of other and unspecified sites

0.512

0.131

127

29,379

0.5610

K91

Postprocedural disorders of digestive system, not elsewhere classified

0.753

0.215

120

29,386

0.0000

K57

Diverticular disease of intestine

0.543

0.196

93

29,413

0.4163

K20

Oesophagitis

0.640

0.066

86

29,420

0.0136

  1. All targets which reached an area under the receiver operating curve (AUC, mean ± standard deviation [std]) of above 0.70 are highlighted in bold. N pos./neg. exams = number of positive/negative examinations (with and without diagnosis, respectively). P-val. P-value for examination scores between groups.
Back to article page