Table 2 Comparison of diagnostic performance for cancer bone metastasis by our and previous AI models.

From: Deep neural network based artificial intelligence assisted diagnosis of bone scintigraphy for cancer bone metastasis

 

EB

BN1

BN2

Ours

Country

Swedish

Japan

Japan

China

Training cohort

795

904

1532

9776

Bone metastases

33%

16%

42%

42%

Age

66 ± 12

64 ± 12

64 ± 12

58 ± 12

Gender

 Male

514

457

790

5421

 Female

281

447

742

4355

Cancer types

 Prostate

431

267

451

1474

 Breast

217

383

624

1680

 Lung

/

/

/

4817

 Others

147

254

457

1805

Validating cohort

NM

NM

NM

1223

Testing cohort

384

257

503

1223

Performance measured by AUC

AUC of cancer types

Prostate

0.939

0.949

0.957

0.955

Breast

0.847

0.91

0.924

0.988

Lung

–

–

–

0.957

Other

0.77

0.861

0.914

0.971

Total

0.858

0.91

0.932

0.964

AUC of gender

 

Male

0.877

0.912

0.934

0.963

Female

0.831

0.91

0.932

0.966

AUC of age

 

 < 60 years

–

–

–

0.979

≥ 60 years

–

–

–

0.949

  1. EB  EXINIbone18, BN1 BONENAVI version 119, BN2 BONENAVI version 220.
Back to article page