Table 1.
Summary of the dataset
|
Train set (70%) (count/total, %) |
Validation set (10%) (count/total, %) |
Test set (20%) (count/total, %) |
Total (count/total, %) |
|
|---|---|---|---|---|
| Labels | ||||
| Infiltration |
13,915/78,484 (17.7%) |
1929/11,211 (17.2%) |
4050/22,425 (18.1%) |
19,894/112,120 (17.7%) |
| Atelectasis |
81,06/78,484 (10.3%) |
1133/11,211 (10.1%) |
2320/22,425 (10.3%) |
11,559/112,120 (10.3%) |
| Effusion |
9401/78,484 (12.0%) |
1315/11,211 (11.7%) |
2601/22,425 (11.6%) |
13,317/11,2120 (11.9%) |
| Nodule |
4392/78,484 (5.6%) |
635/11,211 (5.7%) |
1304/22,425 (5.8%) |
6331/11,2120 (5.6%) |
| Pneumothorax |
3730/78,484 (4.8%) |
520/11,211 (4.6%) |
1052/22,425 (4.7%) |
5302/112,120 (4.7%) |
| Mass |
4016/78,484 (5.1%) |
563/11,211 (5.0%) |
1203/22,425 (5.4%) |
5782/112,120 (5.2%) |
| Consolidation |
3244/78,484 (4.1%) |
480/11,211 (4.3%) |
943/22,425 (4.2%) |
4667/112,120 (4.2%) |
| Pleural_Thickening |
2380/78,484 (3.0%) |
339/11,211 (3.0%) |
666/22,425 (3.0%) |
3385/112,120 (3.0%) |
| Cardiomegaly |
1897/78,484 (2.4%) |
277/11,211 (2.5%) |
602/22,425 (2.7%) |
2776/112,120 (2.5%) |
| Emphysema |
1781/78,484 (2.3%) |
266/11,211 (2.4%) |
469/22,425 (2.1%) |
2516/112,120 (2.2%) |
| Fibrosis |
1204/78,484 (1.5%) |
186/11,211 (1.7%) |
296/22,425 (1.3%) |
1686/112,120 (1.5%) |
| Edema |
1623/78,484 (2.1%) |
244/11,211 (2.2%) |
436/22,425 (1.9%) |
2303/112,120 (2.1%) |
| Pneumonia |
997/78,484 (1.3%) |
128/11,211 (1.1%) |
306/22,425 (1.4%) |
1431/112,120 (1.3%) |
| Hernia |
163/78,484 (0.2%) |
28/11,211 (0.2%) |
36/22,425 (0.2%) |
227/112,120 (0.2%) |
| Count of multi-labels in each image | ||||
| 0 label* |
42,197/78,484 (53.8%) |
6055/11,211 (54.0%) |
12,109/22,425 (54.0%) |
60,361/11,2120 (53.8%) |
| 1 label |
21,768/78,484 (27.7%) |
3103/11,211 (27.7%) |
6092/22,425 (27.2%) |
30,963/112,120 (27.6%) |
| 2 labels |
9993/78,484 (12.7%) |
1427/11,211 (12.7%) |
2886/22,425 (12.9%) |
14,306/112,120 (12.8%) |
| 3 labels |
3361/78,484 (4.3%) |
469/11,211 (4.2%) |
1026/22,425 (4.6%) |
4856/112,120 (4.3%) |
| ≥ 4 labels |
1165/78,484 (1.5%) |
157/11,211 (1.4%) |
312/22,425 (1.4%) |
1634/112,120 (1.5%) |
*0 label indicates “no findings”