Table 4.
Sensitivity of the three radiologists alone and combined with deep learning-generated detections for small and large aneurysms
| Sensitivity for aneurysms < 100 mm3, in % (detected/overall) | Sensitivity for aneurysms > 100 mm3, in % (detected/overall) | |||||
|---|---|---|---|---|---|---|
| Without DLM | With DLM | Difference | Without DLM | With DLM | Difference | |
| Reader 1 | 92.0 (80/87) | 97.7 (85/87) | 5.7 (5/87) | 89.7 (35/39) | 97.4 (38/39) | 7.7 (3/39) |
| Reader 2 | 83.9 (73/87) | 97.7 (85/87) | 13.8 (12/87) | 92.3 (36/39) | 97.4 (38/39) | 5.1 (2/39) |
| Reader 3 | 83.9 (73/87) | 95.4 (83/87) | 11.5 (10/87) | 92.3 (36/39) | 97.4 (38/39) | 5.1 (2/39) |
| Mean | 86.6 | 96.9 | 10.3 | 91.4 | 97.4 | 6.0 |
| 95% CI | 81.3–91.9 | 95.4–98.4 | 5.7–14.9 | 89.7–93.1 | 97.4–97.4 | 4.3–7.7 |
| P-value | 0.024 (vs. without DLM)* | 0.020 (vs. without DLM)* | ||||
Reader 1: 13 years, reader 2: 4 years, and reader 3: 3 years of experience in diagnostic neuroradiology. DLM, deep learning model; CI, confidence interval; SD, standard deviation
*Determined by paired Student’s t-test