Table 2.

Accuracy metrics for lung segmentation from CXR based on U-net applied to two public datasets: NIH³⁵ and Cohen¹⁷.

Metric	NIH dataset		Cohen dataset
	Class 0	Class 1	Class 0	Class 1
Recall	0.981	0.968	0.971	0.919
Precision	0.991	0.934	0.971	0.923
Jaccard	0.973	0.906	0.945	0.854
F1	0.986	0.95	0.971	0.918
Accuracy	0.979	0.979	0.958	0.958

class 0 = background, and class 1 = foreground/lung area.