Table 1 Characteristics of the training, internal validation, external validation, and prospective datasets

From: Automated MRI system for clinically significant prostate cancer detection development validation and real-world implementation

 

Training Set

Validation Set

Test Set. 1

Test Set. 2

Test Set. 3

Test Set. 4

Test Set. 5

Test Set. TCIA

Real-world Set (before ProAI)

Real-world Set (after ProAI)

P-valueb

Cases

4,294

478

246

156

41

51

345

260

927

1,051

 

Age (yr), (Mean ± SD)

64.3 ± 10.7

68.5 ± 8.1

70.0 ± 9.2

68.9 ± 8.6

68.8 ± 8.5

68.1 ± 8.2

66.9 ± 8.4

65.7 ± 7.5

66.4 ± 8.1

66.2 ± 9.3

0.52c

NonPCa (%)

1272(29.6)

189(39.5)

68(27.6)

48(30.8)

14(34.1)

22(43.1)

131(40.0)

67(25.7)

165(17.8)

199(18.9)

0.83 d

NcsPCa (%)

292(6.8)

22(4.6)

23(9.3)

6(3.8)

4(9.8)

6(11.8)

27(7.8)

41(15.8)

18(1.9)

22(2.1)

 

CsPCa (%)

1567(36.5)

267(55.9)

119(48.4)

77(49.4)

12(29.2)

19(37.3)

128(37.1)

152 (58.4)

206(22.2)

241(22.9)

 

MRBx

547(12.7)

0(0)

0(0)

0(0)

0(0)

0(0)

0(0)

0(0)

0(0)

0(0)

 

SysBx

856 (19.9)

126 (26.4)

171 (69.5)

120 (76.9)

29 (70.7)

45 (88.2)

278 (80.6)

252 (96.9)

102(11.0)

138(13.1)

 

SysBx+MRBx

962 (22.4)

240 (50.2)

0(0)

0(0)

0(0)

0(0)

0(0)

8 (3.1)

195(21.0)

237(22.6)

 

RP

766 (17.8)

112 (23.4)

39 (15.9)

11 (7.1)

1 (2.4)

2 (3.9)

8 (2.3)

0(0)

92(9.9)

87(8.3)

 

PI-RADS (%)

         

<0.01 d

 

1

63 (1.5)

11 (2.3)

1 (0.4)

1 (0.6)

0 (0)

2 (3.9)

5 (1.4)

0 (0)

49 (5.3)

71 (6.8)

 

2

2109 (49.1)

108 (22.6)

83 (33.7)

32 (20.5)

18 (43.9)

18 (35.3)

164 (47.5)

73 (22.3)

442 (47.7)

601 (57.2)

 

3

803 (18.7)

85 (17.8)

61 (24.8)

51 (32.7)

6 (14.6)

12 (23.5)

43 (12.5)

85 (25.9)

114 (12.3)

81 (7.7)

 

4

795 (18.5)

133 (27.8)

64 (26.0)

26 (16.7)

4 (9.8)

9 (17.6)

35 (10.1)

111 (33.8)

189 (20.4)

129 (12.3)

 

5

524 (12.2)

141 (29.5)

37 (15.0)

46 (29.5)

13 (31.7)

10 (19.6)

98 (28.4)

59 (18.0)

133 (14.3)

169 (16.1)

 

No. of lesions (%) a

           

0

154(3.6)

9(1.84)

22(8.9)

10(6.4)

1(2.4)

1(1.9)

31(8.9)

6(2.3)

NA

24(2)

 

1

2872(66.9)

340(71.1)

177(71.9)

107(68.5)

34(82.9)

34(66.6)

242(70.1)

156(60)

NA

675(64)

 

2

973(22.7)

104(21.7)

37(15)

22(14.1)

2(4.8)

13(25.4)

56(16.2)

61(23.4)

NA

254(24)

 

3

254(5.9)

23(4.8)

7(2.8)

9(5.7)

2(4.8)

1(1.9)

12(3.4)

30(11.5)

NA

83(8)

 

4

33(0.8)

1(0.2)

2(0.8)

6(3.8)

1(2.4)

2(3.9)

3(0.8)

6(2.3)

NA

14(1)

 

≥5

8(0.2)

1(0.2)

1(0.4)

2(1.2)

1(2.4)

0(0)

1(0.2)

1(0.3)

NA

1(0.01)

 

No. of patients with PSA (%)

3231(75.2)

432(90.4)

230(93.5)

154(98.7)

27(65.9)

50(98.0)

313(90.7)

100(100)

927(100)

1,051 (100)

 

PSA (ng/ml), median (IQR)

9.3 (6.2, 14.1)

9.9 (7.0, 19.7)

10.9 (7.2, 17.0)

12.5 (6.5, 24.8)

11.7 (6.1, 20.8)

12.5 (7.0, 24.0)

12.1 (6.9, 31.4)

7.6 (5.2, 11.8)

9.5(6.4,11.5)

8.7(6.3,12.6)

0.28 d

  1. ISUP International Society of Urology Pathology, non-PCa non-prostate cancer, ncsPCa non-clinically significant prostate cancer (ISUP = 1), csPCa Clinically significant prostate cancer (ISUP > 1), IQR interquartile range, MRBx MRI-guided prostate biopsy, SysBx systematic prostate biopsy, SD standard deviation, RP radical prostatectomy, PSA prostate specific antigen; aAll lesions were generated by AI model. bP-value was compared between group before and after ProAI implementation. cP-value calculated from t-tests. dP-value calculated from Mann-Whitney U test.
Back to article page