Longitudinal patterns in 4-category breast density for the AI density model and interpreting radiologists in the full cohort and in subgroups of women.
| Longitudinal density pattern . | AI density model . | Interpreting radiologists . | P-valuea . |
|---|---|---|---|
| Full cohort (N = 61 177) | |||
| Constant | 49 575 (81.0%) | 34 736 (56.8%) | <.001 |
| Ascending | 1637 (2.7%) | 5065 (8.3%) | |
| Descending | 6953 (11.4%) | 12 021 (19.6%) | |
| Bi-directional | 3012 (4.9%) | 9355 (15.3%) | <.001 |
| Postmenopausal women (N = 27 268) | |||
| Constant | 23 246 (85.3%) | 15 603 (57.2%) | <.001 |
| Ascending | 733 (2.7%) | 2405 (8.8%) | |
| Descending | 2093 (7.7%) | 5018 (18.4%) | |
| Bi-directional | 1196 (4.4%) | 4242 (15.6%) | <.001 |
| Women with stable image-based BMI (N = 4477) | |||
| Constant | 3817 (85.3%) | 2823 (63.1%) | <.001 |
| Ascending | 104 (2.4%) | 356 (8.0%) | |
| Descending | 377 (8.4%) | 887 (19.8%) | |
| Bi-directional | 179 (5.0%) | 411 (9.2%) | <.001 |
| Women with examinations interpreted by the same radiologist (N = 21 216) | |||
| Constant | 17 267 (81.4%) | 14 218 (67.0%) | <.001 |
| Ascending | 584 (2.8%) | 1368 (6.4%) | |
| Descending | 2404 (11.3%) | 3734 (17.6%) | |
| Bi-directional | 961 (4.5%) | 1896 (8.9%) | <.001 |
| Women with examinations interpreted by fellowship-trained radiologists (N = 22 462) | |||
| Constant | 19 696 (87.7%) | 16 122 (71.8%) | <.001 |
| Ascending | 312 (6.4%) | 1286 (5.7%) | |
| Descending | 1446 (1.4%) | 1715 (7.6%) | |
| Bi-directional | 1008 (4.5%) | 3339 (14.9%) | <.001 |
| Longitudinal density pattern . | AI density model . | Interpreting radiologists . | P-valuea . |
|---|---|---|---|
| Full cohort (N = 61 177) | |||
| Constant | 49 575 (81.0%) | 34 736 (56.8%) | <.001 |
| Ascending | 1637 (2.7%) | 5065 (8.3%) | |
| Descending | 6953 (11.4%) | 12 021 (19.6%) | |
| Bi-directional | 3012 (4.9%) | 9355 (15.3%) | <.001 |
| Postmenopausal women (N = 27 268) | |||
| Constant | 23 246 (85.3%) | 15 603 (57.2%) | <.001 |
| Ascending | 733 (2.7%) | 2405 (8.8%) | |
| Descending | 2093 (7.7%) | 5018 (18.4%) | |
| Bi-directional | 1196 (4.4%) | 4242 (15.6%) | <.001 |
| Women with stable image-based BMI (N = 4477) | |||
| Constant | 3817 (85.3%) | 2823 (63.1%) | <.001 |
| Ascending | 104 (2.4%) | 356 (8.0%) | |
| Descending | 377 (8.4%) | 887 (19.8%) | |
| Bi-directional | 179 (5.0%) | 411 (9.2%) | <.001 |
| Women with examinations interpreted by the same radiologist (N = 21 216) | |||
| Constant | 17 267 (81.4%) | 14 218 (67.0%) | <.001 |
| Ascending | 584 (2.8%) | 1368 (6.4%) | |
| Descending | 2404 (11.3%) | 3734 (17.6%) | |
| Bi-directional | 961 (4.5%) | 1896 (8.9%) | <.001 |
| Women with examinations interpreted by fellowship-trained radiologists (N = 22 462) | |||
| Constant | 19 696 (87.7%) | 16 122 (71.8%) | <.001 |
| Ascending | 312 (6.4%) | 1286 (5.7%) | |
| Descending | 1446 (1.4%) | 1715 (7.6%) | |
| Bi-directional | 1008 (4.5%) | 3339 (14.9%) | <.001 |
Data in parentheses are percentages.
P-value for paired proportion hypothesis testing between the AI density model and interpreting radiologists.
Abbreviations: AI = artificial intelligence; BMI = body mass index.
Longitudinal patterns in 4-category breast density for the AI density model and interpreting radiologists in the full cohort and in subgroups of women.
| Longitudinal density pattern . | AI density model . | Interpreting radiologists . | P-valuea . |
|---|---|---|---|
| Full cohort (N = 61 177) | |||
| Constant | 49 575 (81.0%) | 34 736 (56.8%) | <.001 |
| Ascending | 1637 (2.7%) | 5065 (8.3%) | |
| Descending | 6953 (11.4%) | 12 021 (19.6%) | |
| Bi-directional | 3012 (4.9%) | 9355 (15.3%) | <.001 |
| Postmenopausal women (N = 27 268) | |||
| Constant | 23 246 (85.3%) | 15 603 (57.2%) | <.001 |
| Ascending | 733 (2.7%) | 2405 (8.8%) | |
| Descending | 2093 (7.7%) | 5018 (18.4%) | |
| Bi-directional | 1196 (4.4%) | 4242 (15.6%) | <.001 |
| Women with stable image-based BMI (N = 4477) | |||
| Constant | 3817 (85.3%) | 2823 (63.1%) | <.001 |
| Ascending | 104 (2.4%) | 356 (8.0%) | |
| Descending | 377 (8.4%) | 887 (19.8%) | |
| Bi-directional | 179 (5.0%) | 411 (9.2%) | <.001 |
| Women with examinations interpreted by the same radiologist (N = 21 216) | |||
| Constant | 17 267 (81.4%) | 14 218 (67.0%) | <.001 |
| Ascending | 584 (2.8%) | 1368 (6.4%) | |
| Descending | 2404 (11.3%) | 3734 (17.6%) | |
| Bi-directional | 961 (4.5%) | 1896 (8.9%) | <.001 |
| Women with examinations interpreted by fellowship-trained radiologists (N = 22 462) | |||
| Constant | 19 696 (87.7%) | 16 122 (71.8%) | <.001 |
| Ascending | 312 (6.4%) | 1286 (5.7%) | |
| Descending | 1446 (1.4%) | 1715 (7.6%) | |
| Bi-directional | 1008 (4.5%) | 3339 (14.9%) | <.001 |
| Longitudinal density pattern . | AI density model . | Interpreting radiologists . | P-valuea . |
|---|---|---|---|
| Full cohort (N = 61 177) | |||
| Constant | 49 575 (81.0%) | 34 736 (56.8%) | <.001 |
| Ascending | 1637 (2.7%) | 5065 (8.3%) | |
| Descending | 6953 (11.4%) | 12 021 (19.6%) | |
| Bi-directional | 3012 (4.9%) | 9355 (15.3%) | <.001 |
| Postmenopausal women (N = 27 268) | |||
| Constant | 23 246 (85.3%) | 15 603 (57.2%) | <.001 |
| Ascending | 733 (2.7%) | 2405 (8.8%) | |
| Descending | 2093 (7.7%) | 5018 (18.4%) | |
| Bi-directional | 1196 (4.4%) | 4242 (15.6%) | <.001 |
| Women with stable image-based BMI (N = 4477) | |||
| Constant | 3817 (85.3%) | 2823 (63.1%) | <.001 |
| Ascending | 104 (2.4%) | 356 (8.0%) | |
| Descending | 377 (8.4%) | 887 (19.8%) | |
| Bi-directional | 179 (5.0%) | 411 (9.2%) | <.001 |
| Women with examinations interpreted by the same radiologist (N = 21 216) | |||
| Constant | 17 267 (81.4%) | 14 218 (67.0%) | <.001 |
| Ascending | 584 (2.8%) | 1368 (6.4%) | |
| Descending | 2404 (11.3%) | 3734 (17.6%) | |
| Bi-directional | 961 (4.5%) | 1896 (8.9%) | <.001 |
| Women with examinations interpreted by fellowship-trained radiologists (N = 22 462) | |||
| Constant | 19 696 (87.7%) | 16 122 (71.8%) | <.001 |
| Ascending | 312 (6.4%) | 1286 (5.7%) | |
| Descending | 1446 (1.4%) | 1715 (7.6%) | |
| Bi-directional | 1008 (4.5%) | 3339 (14.9%) | <.001 |
Data in parentheses are percentages.
P-value for paired proportion hypothesis testing between the AI density model and interpreting radiologists.
Abbreviations: AI = artificial intelligence; BMI = body mass index.
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
