Table 2 Correlation between AI accuracy and clinical parameters in the P-Model.

From: Diagnostic performance of artificial intelligence for detecting peritoneal and small bowel dissemination in epithelial ovarian cancer using preoperative contrast-enhanced CT imaging

 

Cases without

peritoneal dissemination

n = 131

Cases with

peritoneal dissemination

n = 123

accuracy

 ≤ 25%

accuracy

 > 25%

P-value

accuracy

 ≤ 25%

accuracy

 > 25%

P-value

Variable

 

n = 19

n = 112

 

n = 27

n = 96

 

Ascites volume

on CT, n(%)

The pelvis

12 (63.2)

106 (94.6)

 < 0.001*

19 (70.4)

42 (43.8)

0.017*

The upper abdomen

7 (36.8)

6 (5.4)

 

8 (29.6)

54 (56.2)

 

Intraoperative ascites(ml), n(%)

 ≤ 300

15 (78.9)

104 (92.9)

0.074

21 (77.8)

41 (42.7)

0.002*

300 < 

4 (21.1)

8 (7.1)

 

6 (22.2)

55 (57.3)

 

BMI(kg/m2), mean [range]

 

22.7

[16.6- 30.1]

22.0

[15.4- 34.3]

0.27

22.1

[17.1- 35.2]

21.9

[14.6- 32.9]

0.77

CA125(U/ml), mean [range]

 

1165.2

[13- 10769]

549.7

[4- 15882]

0.005*

1437.9

[2- 23961]

4686.1

[9- 142522]

 < 0.001*

PCI score, mean [range]

 

0

[0- 0]

0.12

[0- 8]

0.47

8.46

[1-24]

12.98

[1- 36]

0.005*

Histological subtype, n(%)

HGSC

4 (21.1)

22 (19.6)

1

18 (66.7)

79 (82.3)

0.11

Non- HGSC

15 (78.9)

90 (80.4)

 

9 (33.3)

17(17.7)