Skip to main content
. 2021 Jan 16;113:107828. doi: 10.1016/j.patcog.2021.107828

Table 2.

Quantitative comparison for severity assessment tasks with the state-of-the-art methods.

Method Accuracy Precision Recall F1 Score AUC
ResNet50-3d 0.913±0.059 0.902±0.118 0.756±0.106 0.786±0.112 0.759±0.102
ResNet50+Max 0.924±0.497 0.856 ± 0.154 0.793±0.106 0.816± 0.120 0.803±0.090
Gated Att. MIL[39] 0.955±0.015 0.879±0.054 0.946±0.019 0.906± 0.037 0.973±0.024
Tang et al.[3]* 0.875 - 0.933 - 0.910
Yang et al.[37] - - 0.750 - 0.892
Cls. Only (Ours) 0.969±0.023 0.928±0.073 0.958±0.031 0.938±0.045 0.980±0.013
M2UNet (Ours) 0.985±0.005 0.975±0.022 0.952±0.011 0.963±0.011 0.991±0.010