Skip to main content

Table 2 Demographics distributions for the corpus under study in terms of images

From: A radiographic, deep transfer learning framework, adapted to estimate lung opacities from chest x-rays

Images

AllData

Train

Test

Validation

OOB Test

Image Count N (%)

38,365 (100)

30,586 (79.72)

3660 (9.53)

3832 (9.99)

286 (0.74)

Xray view

 PA (%)

456 (1.18)

374 (1.22)

39 (1.06)

40 (1.04)

3 (1.04)

 AP (%)

36,596 (95.38)

29,165 (95.35)

3491 (95.38)

3665 (95.61)

275 (96.15)

 Unknown (%)

1313 (3.42)

1047 (3.42)

130 (3.55)

128 (3.33)

8 (2.79)

CXR opacity - left

 None

12,488 (32.55)

9983 (32.63)

1234 (33.71)

1174 (30.62)

97 (33.91)

 Mild

12,381 (32.27)

9798 (32.03)

1189 (32.48)

1300 (33.91)

94 (32.86)

 moderate

8310 (21.66)

6625 (21.66)

760 (20.76)

862 (22.48)

63 (22.02)

 severe

5185 (13.51)

4180 (13.66)

477 (13.03)

496 (12.94)

32 (11.18)

CXR opacity - right

 None

12,938 (33.72)

10,237 (33.46)

1297 (35.43)

1311 (34.21)

93 (32.51)

 mild

11,382 (29.66)

9080 (29.68)

1051 (28.71)

1171 (30.55)

80 (27.97)

 moderate

8459 (22.04)

6771 (22.13)

792 (21.63)

821 (21.42)

75 (26.22)

 severe

5585 (14.55)

4498 (14.70)

520 (14.20)

529 (13.80)

38 (13.28)

  1. Image-wise demographics for CXR view and opacity degree.