Table 2.

Comparison of classification performance on individual RBCs using Mask R-CNN, XGBoost with manual delineation and XGBoost with Mask R-CNN segmentation results.

Method	Sensitivity (%)	Specificity (%)	$F1$-score	AUC	Accuracy (%)
Mask R-CNN	97.1	98.5	0.975	0.982	97.8
XGBoost with MD	99.9	100	0.992	0.999	99.9
XGBoost with Mask R-CNNa	99.9	100	0.992	0.999	99.9

Note: AUC, area under the receiver operating characteristic curve; MD, manual delineation.

^aXGBoost classifier trained with the segmentation results from Mask R-CNN.