Table 4 Performance of screening mammography compared between radiologists and standalone AI according to AI-based breast density
Outcome | Radiologists’ BI-RADS category (0, 3, 4, 5) | Standalone AI (Cutoff 10%) | P value | ||
|---|---|---|---|---|---|
Estimate | 95% CI | Estimate | 95% CI | ||
Non-dense (A or B) | |||||
CDR, per 1000 examinations | 1.2 | 0.7–1.8 | 1.0 | 0.7–1.7 | 0.317 |
Sensitivity, % | 83.3 | 62.6–95.3 | 75 | 53.3–90.2 | 0.317 |
Specificity, % | 78.8 | 78.1–79.4 | 96.3 | 96.0–96.6 | < 0.001 |
PPV, % | 0.5 | 0.3–0.8 | 2.8 | 1.6–4.3 | < 0.001 |
Recall rate, % | 21.3 | 20.7–22.0 | 3.8 | 3.5–4.1 | < 0.001 |
AUC | 0.81 | 0.73–0.89 | 0.86 | 0.77–0.95 | 0.268 |
Heterogeneously dense (C) | |||||
CDR, per 1000 examinations | 1.1 | 0.9–1.4 | 1.0 | 0.8–1.3 | 0.257 |
Sensitivity, % | 69.2 | 58.7–78.5 | 62.6 | 51.9–72.6 | 0.257 |
Specificity, % | 77.1 | 76.7–77.4 | 93.2 | 92.9–93.4 | < 0.001 |
PPV, % | 0.5 | 0.4–0.6 | 1.5 | 1.1–1.9 | < 0.001 |
Recall rate, % | 23.0 | 22.7–23.4 | 6.9 | 6.7–7.2 | < 0.001 |
AUC | 0.73 | 0.68–0.78 | 0.78 | 0.73–0.83 | 0.103 |
Extremely dense (D) | |||||
CDR, per 1000 examinations | 1 | 0.6–1.6 | 1.3 | 0.8–1.9 | 0.103 |
Sensitivity, % | 60.7 | 40.6–78.5 | 75.0 | 55.1–89.3 | 0.103 |
Specificity, % | 78.3 | 77.7–78.9 | 89.2 | 88.8–89.7 | < 0.001 |
PPV, % | 0.5 | 0.3–0.8 | 1.2 | 0.7–1.8 | 0.004 |
Recall rate, % | 21.8 | 21.1–22.4 | 10.9 | 10.4–11.3 | < 0.001 |
AUC | 0.70 | 0.60–0.79 | 0.82 | 0.74–0.90 | 0.003 |