Table 1 Characteristics of the study sample.

From: A joint convolutional-recurrent neural network with an attention mechanism for detecting intracranial hemorrhage on noncontrast head CT

Variables

Study sample

(n = 48,070)

Training

(n = 43,460)

Validation

(n = 4610)

Testing

(n = 380)

Age

54 (IQR, 43–65)

54 (IQR, 43–65)

53 (IQR, 40–60)

48 (IQR, 36–57)

Gender (Male)

28,253 (58.77%)

25,079 (57.70%)

3174 (68.85%)

 

ICH-positive patients

13,224 (27.50%)

9226 (21.22%)

612 (13.27%)

130 (34.21%)

CT examinations

55,179

49,968

5211

452

ICH (binary)

15,733 (28.51%)

14,742 (29.50%)

991 (19.01%)

167 (36.9%)

IPH

10,080 (18.26%)

9422 (18.85%)

658 (12.62%)

86 (19.02%)

IVH

5963 (10.78%)

5535 (11.07%)

418 (8.02%)

38 (8.4%)

SAH

9555 (17.31%)

8955 (17.92%)

600 (11.51%)

48 (10.61%)

SDH

7473 (13.54%)

7022 (14.05%)

451 (8.65%)

76 (16.81%)

EDH

1237 (2.24%)

1116 (2.33%)

71 (1.35%)

14 (3.1%)

Total CT slices

2,255,271

2,255,271

212,873

ICH-positive

188,067 (8.33%)

175,664 (7.78%)

12,403 (5.82%)

ICH-negative

2,067,204 (91.67%)

2,079,607 (92.22%)

200.470 (6.18%)

  1. *EDH epidural hemorrhage, ICH intracranial hemorrhage, IPH intra-parenchymal hemorrhage, SAH subarachnoid hemorrhage, SDH subdural hemorrhage.