Table 7. The performance of QUARTZ on the two datasets FOREVERP and UK Biobank.
Performance evaluated on the following parameters: Vessel segmentation, arteriole, and venule classification. Performance parameters are addressed as sensitivity, specificity, and accuracy. Vessel segmentation reported with and without pre/postprocessing. Arteriole and venule classification reported per pixel and per segment.
| Dataset Parameter |
FORVEVERP | UK Biobank | |||||
|---|---|---|---|---|---|---|---|
| Sensitivity | Specificity | Accuracy | Sensitivity | Specificity | Accuracy | Comment | |
| Vessel segmentation |
80.31% | 98.02% | 96.44% | 73.66% | 98.14% | 95.64% | without pre/post-processing |
| 74.64% | 98.41% | 96.28% | 69.12% | 98.88% | 95.84% | with pre/post-processing | |
| Arteriole classification | 90.23 (A) | 88.36 (A) | 89.25 (A) | 86.07 (A) | 87.67 (A) | 86.97(A) | level: pixel |
| 88.64 (A) | 84.26 (A) | 86.32 (A) | 85.14 (A) | 85.32 (A) | 85.24 (A) | level: segment | |
| Venule classification | 88.36 (V) | 90.23 (V) | 89.25 (V) | 87.67 (V) | 86.07 (V) | 86.97 (V) | level: pixel |
| 84.26 (V) | 88.64 (V) | 86.32 (V) | 85.32 (V) | 85.14 (V) | 85.24 (V) | level: segment | |