Table 3 Comparison of clinical characteristics between deep-learning set and validation set.

From: Predicting multiple linear stapler firings in double stapling technique with an MRI-based deep-learning model

 

Deep-learning set

Validation set

p value

n = 328

n = 128

Baseline characteristics

 Age [y]

  

0.139

  > 70

70 (21.3)

36 (28.1)

 

  ≤ 70

258 (78.0)

92 (71.9)

 

 Gender, n (%)

  

0.561

  Male

227 (69.2)

83 (64.8)

 

  Female

101 (30.8)

45 (35.2)

 

 BMI [kg/m2]

  

0.250

  > 25

89 (27.1)

42 (32.8)

 

  ≤ 25

239 (72.9)

86 (67.2)

 

 Diabetes mellitus, n (%)

  

0.342

  Yes

41 (12.5)

23 (18.0)

 

  No

287 (87.5)

105 (82.0)

 

 nCRT, n (%)

  

0.201

  Yes

85 (25.9)

41 (32.0)

 

  No

243 (74.1)

87 (68.0)

 

 Number of linear stapler cartridges, n (%)

  

0.204

  ≥ 3

58 (17.7)

16 (12.5)

 

  ≤ 2

270 (82.3)

112 (87.5)

 

 Anastomotic leakage, n (%)

  

0.090

  Yes

48 (14.6)

11 (8.6)

 

  No

280 (85.4)

117 (91.4)

 

Biochemical data

 Albumin [g/L]

  

0.356

  < 35

97 (29.6)

32 (25.0)

 

  ≥ 35

231 (70.4)

96 (75.0)

 

 CEA [ng/mL], n (%)

(Missing = 5)

 

0.301

  > 5

97 (29.6)

32 (25.0)

 

  ≤ 5

226 (68.9)

96 (75.0)

 

Tumor characteristics

 Distance from anus [cm]

  

0.637

  < 5

43 (13.1)

14 (10.9)

 

  ≥ 5

285 (86.9)

114 (89.1)

 

 Tumor size [cm], n (%)

  

0.130

  ≥ 5

40 (12.2)

9 (7.0)

 

  < 5

288 (87.8)

119 (93.0)

 

 CRM (evaluated by MRI), n (%)

  

0.092

  Positive

72 (22.0)

19 (14.8)

 

  Negative

256 (78.0)

109 (69.5)

 
  1. MRI: magnetic resonance imaging; CRM: circumferential resection margin; CEA: carcinoma embryonic antigen; nCRT: neoadjuvant chemoradiotherapy; BMI: body mass index.