Table 2.
Characteristics of the training, validation and test sets
| Variable | Training | Validation | Test | |
|---|---|---|---|---|
| Cases | Participants | 11,832 | 1912 | 9313 |
| Studies | 36,093 | 5982 | 27,622 | |
| Images | 230,954 | 39,494 | 181,194 | |
| BI-RADS density | A | 4132 (11) | 720 (12) | 3093 (11) |
| B | 15,433 (43) | 2660 (44) | 11,009 (40) | |
| C | 14,749 (41) | 2266 (38) | 11,710 (42) | |
| D | 1779 (5) | 336 (6) | 1810 (7) | |
| Imaging technique | FFDM | 19,323 (53) | 2929 (49) | 13,335 (48) |
| FFDM + 2DS | 16,770 (47) | 3053 (51) | 14,287 (52) | |
| Race | Black | 19,152 (53) | 3404 (57) | 13,945 (50) |
| White | 16,941 (47) | 2578 (43) | 10,093 (37) | |
| Asian | 0 (0) | 0 (0) | 3584 (13) | |
Breakdown for BI-RADS density, imaging technique, and races is given as the number of studies. Percentages are given in brackets.
FFDM full-field digital mammography, 2DS two-dimensional synthetic.