Skip to main content
. 2020 Oct 29;25(5):1735–1746. doi: 10.1109/JBHI.2020.3034863

TABLE III. Detection Accuracy for Our RBC Point Set. All Training Data of the Polygon Set (165 Images) has Been Used to Generate the Training Model to Test 800 Images From 160 Patients. The t-Tests Between Our Proposed Dual RBCNet and Other Methods Have P-Values < 0.001, Indicating That the Differences Between Groups are Statistically Significant.

Method Inline graphic Evaluation Metrics F1-Measure Inline graphic Precision Inline graphic Recall Inline graphic
Traditional methods
Watershed [17] Inline graphic Inline graphic Inline graphic
Active contour[18], [19] Inline graphic Inline graphic Inline graphic
Instance segmentation deep learning methods
SegNet [48] Inline graphic Inline graphic Inline graphic
U-Net [42] Inline graphic Inline graphic Inline graphic
DeepLab v3+ [49] Inline graphic Inline graphic Inline graphic
Object detection deep learning methods
Faster R-CNN [39] on overlapping-tiles + NMS Inline graphic Inline graphic Inline graphic
Yolo [50] on overlapping-tiles + NMS Inline graphic Inline graphic Inline graphic
SSD [51] on overlapping-tiles + NMS Inline graphic Inline graphic Inline graphic
Mask R-CNN [52] on overlapping-tiles + NMS Inline graphic Inline graphic Inline graphic
Proposed dual deep learning networks
SegNet + Faster R-CNN Inline graphic Inline graphic Inline graphic
U-Net + YOLO Inline graphic Inline graphic Inline graphic
RBCNet (U-Net + Faster R-CNN) Inline graphic Inline graphic Inline graphic