Table 1.
Characteristics of the study population.
| Attribute | CheXpert |
MIMIC-CXR |
||||||
|---|---|---|---|---|---|---|---|---|
| All | White | Asian | Black | All | White | Asian | Black | |
| All data | ||||||||
| Patients | 42,884 | 33,338 | 6642 | 2904 | 43,209 | 32,756 | 1881 | 8572 |
| Scans | 127,118 | 99,027 (78) | 18,830 (15) | 9261 (7) | 183,207 | 14,1865 (77) | 7106 (4) | 34,236 (19) |
| Age (years) | 63 ± 17 | 64 ± 17 | 61 ± 17 | 56 ± 17 | 65 ± 17 | 66 ± 16 | 63 ± 18 | 61 ± 17 |
| Female | 52,436 (41) | 39,735 (40) | 8132 (43) | 4569 (49) | 85,193 (47) | 61,626 (43) | 31,22 (44) | 20,445 (60) |
| No finding | 10,916 (9) | 8236 (8) | 1716 (9) | 964 (10) | 56,615 (31) | 41,215 (29) | 22,21 (31) | 13,179 (38) |
| Pleural effusion | 51,574 (41) | 40,545 (41) | 7953 (42) | 3076 (33) | 46,224 (25) | 38,693 (27) | 19,16 (27) | 5615 (16) |
| Training data | ||||||||
| Patients | 25,730 | 20,034 | 3945 | 1751 | 25,925 | 19,613 (76) | 1110 (4) | 5202 (20) |
| Scans | 76,205 | 59,238 (78) | 11,371 (15) | 5596 (7) | 110,280 | 86,098 (78) | 4248 (4) | 19,934 (18) |
| Age (years) | 63 ± 17 | 64 ± 17 | 62 ± 17 | 56 ± 17 | 65 ± 17 | 66 ± 16 | 63 ± 18 | 60 ± 17 |
| Female | 31,432 (41) | 23,715 (40) | 4976 (44) | 2741 (49) | 51,138 (46) | 37,518 (44) | 1897 (45) | 11,723 (59) |
| No finding | 6514 (9) | 4910 (8) | 1046 (9) | 558 (10) | 34,530 (31) | 25,170 (29) | 1330 (31) | 8030 (40) |
| Pleural effusion | 31,015 (41) | 24,405 (41) | 4754 (42) | 1856 (33) | 27,806 (25) | 23,526 (27) | 11,08 (26) | 3172 (16) |
| Validation data | ||||||||
| Patients | 4288 | 3348 | 666 | 274 | 4321 | 3242 (75) | 209 (5) | 870 (20) |
| Scans | 12,673 | 9945 (79) | 1809 (14) | 919 (7) | 17,665 | 13,369 (76) | 776 (4) | 3520 (20) |
| Age (years) | 62 ± 17 | 63 ± 17 | 62 ± 17 | 55 ± 16 | 65 ± 17 | 67 ± 16 | 60 ± 22 | 62 ± 17 |
| Female | 5030 (40) | 3933 (40) | 667 (37) | 430 (47) | 8245 (47) | 5755 (43) | 336 (43) | 2154 (61) |
| No finding | 1086 (9) | 817 (8) | 175 (10) | 94 (10) | 5393 (31) | 3903 (29) | 232 (30) | 1258 (36) |
| Pleural effusion | 5049 (40) | 3988 (40) | 738 (41) | 323 (35) | 4575 (26) | 3721 (28) | 230 (30) | 624 (18) |
| Test data | ||||||||
| Patients | 12,866 | 9956 | 2031 | 879 | 12,963 | 9901 (76) | 562 (5) | 2500 (19) |
| Scans | 38,240 | 29,844 (78) | 5650 (15) | 2746 (7) | 55,262 | 42,398 (77) | 2082 (4) | 10,782 (19) |
| Age (years) | 63 ± 17 | 64 ± 17 | 61 ± 17 | 57 ± 16 | 65 ± 17 | 66 ± 16 | 65 ± 17 | 61 ± 17 |
| Female | 15,974 (42) | 12,087 (41) | 2489 (44) | 1348 (49) | 25,810 (47) | 18,353 (43) | 889 (43) | 6568 (61) |
| No finding | 3316 (9) | 2509 (8) | 495 (9) | 312 (11) | 16,692 (30) | 12,142 (29) | 659 (32) | 3891 (36) |
| Pleural effusion | 15,510 (41) | 12,152 (41) | 2461 (44) | 897 (33) | 13,843 (25) | 11,446 (27) | 578 (28) | 1819 (17) |
Breakdown of demographics over the set of patient scans by racial groups and training, validation and test splits. Percentages in brackets are with respect to the number of scans. We also report the number of unique patients for each group.