Table 2 Quantitative metrics comparisons between cell-level and coarse labels for HybridSN-Att direct predictions.

From: Improving lung cancer pathological hyperspectral diagnosis through cell-level annotation refinement

Label

Accuracy(%)

Specificity(%)

Sensitivity(%)

Dice(%)

Coarse

77.33(±7.96)

84.68(±6.45)

66.09(±13.61)

60.60(±13.30)

Cell-level

92.52(±4.36)

97.26(±1.66)

54.34(±17.21)

55.28(±13.14)

  1. Significant values are in [bold]
Back to article page