Table 1 Comparing the average AUC scores (with a 95% confidence interval) of the PE classification from CXR, CTPA and Generated CTPAs on our test set
From: X-ray2CTPA: leveraging diffusion models to enhance pulmonary embolism classification
Model | AUC | Accuracy | Specificity | Sensitivity | PPV | NPV |
|---|---|---|---|---|---|---|
Baselines | ||||||
CTPA only classifier | 0.90 [0.88–0.97] ± 0.04 | 84.05 | 83.86 | 82.74 | 69.78 | 92.32 |
CXR only classifier | 0.691 [0.54–0.77] ± 0.06 | 71.79 | 70.91 | 73.91 | 51.52 | 86.67 |
Generated Data | ||||||
Generated CTPA only classifier | 0.80 [0.72–0.9] ± 0.07 | 74.0 | 74.1 | 73.91 | 54.8 | 87.0 |
