Performance of the occlusion tools
Totally, 149 attempts of occlusion in 92 PVs were recorded and analyzed. Figure 2 shows the occlusion assessment results of different PVs when using the baseline, injection, and pulmonary venography as verification tools. Among PVs, optimal occlusion could be frequently achieved in LSPV and RSPV, whereas moderate to severe leak was commonly detected in LIPV and RIPV.
Using pulmonary venography as the gold standard, the overall sensitivity, specificity, PPV, NPV of the baseline tool was 96.7% (95% CI 90.0% - 99.1%), 40.5% (95% CI 26.0% - 56.7%), 77.9% (95% CI 68.9% - 84.9%), and 85.0% (95% CI 61.1% - 96.0%), respectively, while the corresponding value of the injection tool was 69.6% (95% CI 59.7% - 78.1%), 100.0% (95% CI 90.6% - 100.0%), 100.0% (95% CI 93.6% - 100.0%), and 60.3% (95% CI 48.5% - 70.0%), respectively. As shown in ROC curves, the AUC of the baseline and injection tool was 0.693 (95% CI 58.8% - 79.8%) and 0.848 (78.8% - 90.8%), respectively (Figure 3). Furthermore, the sensitivity, specificity, PPV, NPV of the 2 tools for different PVs were listed in table 2.