Table 1 Patient characteristics for the Stanford-HCC and TCGA-HCC datasets.

From: Deep learning predicts postsurgical recurrence of hepatocellular carcinoma from digital histopathologic images

Patient characteristic

TCGA-HCC development cohort (n = 299)

TCGA-HCC test cohort (n = 53)

Stanford-HCC (n = 198)

Training and validation set

Internal test set

External test set

Age (at surgery) (years)

60 (51, 68)

61 (51, 68)

64 (57, 69)

Gender

Male

206 (69%)

33 (62%)

157 (79%)

Female

93 (31%)

20 (38%)

41 (21%)

Hepatitis B virus infection

Negative

195 (68%)

33 (67%)

147 (74%)

Positive

90 (32%)

16 (33%)

51 (26%)

Unknown

14

4

0

Hepatitis C virus infection

Negative

243 (85%)

41 (84%)

96 (48%)

Positive

42 (15%)

8 (16%)

102 (52%)

Unknown

14

4

0

Alcohol intake

Negative

188 (66%)

30 (61%)

181 (91%)

Positive

97 (34%)

19 (39%)

17 (8.6%)

Unknown

14

4

0

Non-alcoholic fatty liver disease

Negative

271 (95.1%)

44 (90%)

184 (93%)

Positive

14 (4.9%)

5 (10%)

14 (7.1%)

Unknown

14

4

0

AJCC stage grouping

IA

8 (2.7%)

1 (1.9%)

44 (22%)

IB

121 (41%)

24 (46%)

66 (33%)

II

85 (29%)

16 (31%)

68 (34%)

IIIA

60 (20%)

9 (17%)

11 (5.6%)

IIIB

16 (5.4%)

1 (1.9%)

7 (3.5%)

IVA

3 (1.0%)

0 (0%)

2 (1%)

IVB

1 (0.3%)

1 (1.9%)

0 (0%)

Unknown

5

1

0

Largest tumor diameter (mm)

65 (35, 100)

55 (34, 100)

30 (18, 50)

Unknown

5

0

0

Tumor multifocality

Negative

207 (69%)

38 (73%)

142 (72%)

Positive

91 (31%)

14 (27%)

56 (28%)

Unknown

1

1

0

Histologic grade

Well-differentiated

46 (15%)

5 (9.4%)

63 (32%)

Moderately-differentiated

162 (54%)

32 (60%)

108 (55%)

Poorly-differentiated

89 (30%)

16 (30%)

26 (13%)

Undifferentiated

2 (0.7%)

0 (0%)

1 (0.5%)

Microvascular invasion

Negative

196 (67%)

36 (68%)

147 (74%)

Positive

95 (33%)

17 (32%)

51 (26%)

Unknown

8

0

0

Macrovascular invasion

Negative

271 (93%)

49 (92%)

188 (96%)

Positive

21 (7.2%)

4 (7.5%)

8 (4.1%)

Unknown

7

0

0

Surgical margin status

Negative

249 (94%)

45 (88%)

192 (97%)

Positive

17 (6.4%)

6 (12%)

5 (2.5%)

Unknown

33

2

0

Fibrosis stage

0

77 (33%)

13 (30%)

38 (19%)

1

11 (4.8%)

2 (4.5%)

13 (6.6%)

2

25 (11%)

4 (9.1%)

15 (7.6%)

3

26 (11%)

6 (14%)

13 (6.6%)

4

91 (40%)

19 (43%)

119 (60%)

Unknown

69

9

0

Recurrence

No

148 (49%)

28 (53%)

136 (69%)

Yes

151 (51%)

25 (47%)

62 (31%)

Length of follow-up (months)

12 (4, 24)

13 (6, 20)

25 (9, 48)

Risk score

 

0.07 (− 0.26, 0.30)

− 0.31 (− 0.46, − 0.15)

  1. Values presented: median (IQR); n (%).