Table 2.
Selection of best mean classification accuracies achieved for prediction of each assessed imaging biomarker.
| Radiomics Analysis to Predict | Best Results by | AUC | Sensitivity | Specificity | PPV | NPV | Accuracy |
|---|---|---|---|---|---|---|---|
| Subtype (luminal A versus luminal B) |
All MR | 0.978 (0.950–1.000) |
94.6 (87.9–98.2) |
100.0 (96.0–100.0) |
100.0 (95.9–100.0) |
94.7 (88.1–98.3) |
97.3 (93.7–99.1) |
| Subtype (luminals vs. others) | PET | 0.950 (0.922–0.979) |
83.5 (74.6–90.3) |
93.2 (86.5–97.2) |
92.0 (84.3–96.7) |
85.7 (77.8–91.6) |
88.5 (83.2–92.6) |
| ER Status (negative vs. positive) | All MR and PET | 0.870 (0.818–0.923) |
90.1 (82.1–95.4) |
65.9 (55.0–75.7) |
73.2 (64.0 -81.1) |
86.6 (76.0–93.7) |
78.2 (71.4–84.0) |
| PR Status (negative vs. positive) | All MR and PET | 0.879 (0.826–0.932) |
84.1 (74.8–91.0) |
83.9 (74.8–90.7) |
83.1 (73.7–90.2) |
84.8 (75.8–91.4) |
84.0 (77.8–89.0) |
| HER2 (negative vs. positive) |
All DCE | 0.972 (0.955–0.989) |
84.9 (76.6–91.1) |
93.2 (86.5–97.2) |
92.8 (85.7–97.0) |
85.7 (77.8–91.6) |
89.0 (83.9–92.9) |
| Proliferation (high vs. low) |
All MR and PET | 0.997 (0.992–1.000) |
99.1 (95.1–100.0) |
92.7 (86.0–96.8) |
93.2 (87.1–97.0) |
99.0 (94.7–100.0) |
95.9 (92.4–98.1) |
| Grade (grade 1 vs. grade 2 vs. grade 3) | PET | 0.771 (0.693–0.849) |
66.2 (53.7–77.2) |
78.1 (66.9–86.9) |
73.8 (60.9–84.2) |
71.3 (60.0–80.8) |
72.3 (64.2–79.5) |
| Nodal Status (0 vs. 1, 2, 3) | All MR and PET | 0.810 (0.740–0.881) |
63.8 (51.3–75.0) |
82.2 (71.5–90.2) |
77.2 (64.2–87.3) |
70.6 (59.7–80.0) |
73.2 (65.2–80.3) |
| Distant Metastases (0 vs. 1) | All MR and PET | 0.999 0.997–1.000) |
98.3 (94.0–99.8) |
98.3 (94.0–99.8) |
98.3 (94.0–99.8) |
98.3 (94.0–99.8) |
98.3 (95.7–99.5) |