Table. 1 Baseline clinicopathological characteristics of the training and validation set and internal test set of surgical cohorts (n = 160)

From: Artificial intelligence-driven prediction of lymph node metastasis in T1 esophageal squamous cell carcinoma using whole slide images

 

Training and validation set Surgical cohort (N = 112)

Internal test set Surgical cohort (N = 48)

P-value

Age (years)

66.77 ± 7.752

67.42 ± 7.234

0.612

Sex, n (%)

  

0.216

 Male

86 (76.8%)

41 (85.4%)

 

 Female

26 (23.2%)

7 (14.6%)

 

Tumor size (mm)

39.32 ± 15.15

31.54 ± 13.61

0.265

Location, n (%)

  

0.961

 Upper

7 (6.3%)

3 (6.3%)

 

 Middle

54 (48.2%)

22 (45.8%)

 

 Lower

51 (45.5%)

23 (47.9%)

 

Invasion depth

  

0.896

 Submucosa

15 (13.4%)

7 (14.6%)

 

 Muscularis propria

25 (22.3%)

12 (25.0%)

 

 Adventitia

72 (64.3%)

29 (60.4%)

 

Clinical stage

  

0.908

 Stage Ⅰ

15 (13.4%)

8 (16.7%)

 

 Stage Ⅱ

53 (47.3%)

20 (41.7%)

 

 Stage Ⅲ

40 (35.7%)

18 (37.5%)

 

 Stage IV

4 (3.6%)

2 (4.2%)

 

Lymphovascular invasion, n (%)

69 (61.6%)

31 (64.6%)

0.722

Perineural invasion, n (%)

51 (45.5%)

25 (52.1%)

0.447

Differentiation grade, n (%)

  

0.405

 Poorly differentiated

20 (17.9%)

13 (27.1%)

 

 Moderately differentiated

87 (77.7%)

34 (70.8%)

 

 Well differentiated

5 (4.5%)

1 (2.1%)

 

Tumor budding*

  

0.405

 Low-grade budding

5 (4.5%)

1 (2.1%)

 

 Intermediate-grade budding

87 (77.7%)

34 (70.8%)

 

 High-grade budding

20 (17.9%)

13 (27.1%)

 

Lymph node metastasis (+), n (%)

52 (46.4%)

20 (41.6%)

0.579

Number of dissected lymph nodes, n

29.98 ± 11.21

27.65 ± 10.50

0.426

  1. *Tumor budding, defined as isolated cancer cell clusters (≤5 cells) at the invasive margin, was graded as BD1 (0–4 buds/field), BD2 (5–9 buds/field), or BD3 (≥10 buds/field) under 200× magnification.