Table 3 The performance of DL model in predicting MIBC at different VI-RADS.

From: Deep learning on T2WI to predict the muscle-invasive bladder cancer: a multi-center clinical study

VI-RADS

Groups

Accuracy

% (n/N)

Sensitivity

% (n/N)

Specificity

% (n/N)

PPV

% (n/N)

NPV

% (n/N)

VI-RADS 1

(n = 21)

All patients (n = 21)

100 (21/21)

N/A

100 (21/21)

N/A

100 (21/21)

Validation set (n = 6)

100 (6/6)

N/A

100 (6/6)

N/A

100 (6/6)

Internal test set (n = 13)

100 (13/13)

N/A

100 (13/13)

N/A

100 (13/13)

External test set (n = 2)

100 (2/2)

N/A

100 (2/2)

N/A

100 (2/2)

VI-RADS 2

(n = 123)

All patients (n = 123)

93.5 (115/123)

100 (1/1)

93.4 (114/122)

11.1 (1/9)

100 (114/114)

Validation set (n = 33)

90.9 (30/33)

N/A

90.9 (30/33)

N/A

100 (30/30)

Internal test set (n = 67)

95.5 (64/67)

100 (1/1)

95.5 (63/66)

25.0 (1/4)

100 (63/63)

External test set (n = 23)

91.3 (21/23)

N/A

91.3 (21/23)

N/A

100 (21/21)

VI-RADS 3

(n = 60)

All patients (n = 60)

80.0 (48/60)

66.7 (14/21)

87.2 (34/39)

73.7 (14/19)

82.9 (34/41)

Validation set (n = 11)

81.8 (9/11)

75.0 (3/4)

85.7 (6/7)

75.0 (3/4)

85.7 (6/7)

Internal test set (n = 42)

83.3 (35/42)

68.8 (11/16)

92.3 (24/26)

84.6 (11/13)

82.8 (24/29)

External test set (n = 7)

57.1 (4/7)

0 (0/1)

66.7 (4/6)

0 (0/2)

80.0 (4/5)

VI-RADS 4

(n = 31)

All patients (n = 31)

90.3 (28/31)

91.7 (22/24)

85.7 (6/7)

95.7 (22/23)

75.0 (6/8)

Validation set (n = 8)

100.0 (8/8)

100.0 (7/7)

100.0 (1/1)

100.0 (7/7)

100.0 (1/1)

Internal test set (n = 21)

90.5 (19/21)

93.3 (14/15)

83.3 (5/6)

93.3 (14/15)

83.3 (5/6)

External test set (n = 2)

50.0 (1/2)

50.0 (1/2)

N/A

100 (1/1)

0 (0/1)

VI-RADS 5

(n = 33)

All patients (n = 33)

93.9 (31/33)

93.9 (31/33)

N/A

100 (31/31)

N/A

Validation set (n = 8)

100.0 (8/8)

100.0 (8/8)

N/A

100.0 (8/8)

N/A

Internal test set (n = 21)

95.2 (20/21)

95.2 (20/21)

N/A

100.0 (20/20)

N/A

External test set (n = 4)

75.0 (3/4)

75.0 (3/4)

N/A

100.0 (3/3)

N/A