Table 3 The performance metrics of the two models
From: Effectiveness of opportunistic osteoporosis screening on chest CT using the DCNN model
Model 1 | Model 2 | ||||||
|---|---|---|---|---|---|---|---|
Training dataset | Validation dataset | Test dataset | Training dataset | Validation dataset | Test dataset | ||
normal | AUC | 0.999 | 0.972 | 0.989 | 0.999 | 0.966 | 0.983 |
(95%CI) | (0.998, 1.000) | (0.939, 1.000) | (0.983, 0.996) | (0.998, 1.000) | (0.924, 1.000) | (0.974, 0.992) | |
Se | 0.990 | 0.917 | 0.964 | 1.000 | 1.000 | 0.976 | |
Sp | 0.972 | 0.900 | 0.916 | 0.927 | 0.850 | 0.838 | |
PPV | 0.981 | 0.932 | 0.945 | 0.951 | 0.909 | 0.901 | |
NPV | 0.986 | 0.878 | 0.944 | 1.000 | 1.000 | 0.959 | |
Ac | 0.983 | 0.910 | 0.945 | 0.969 | 0.940 | 0.921 | |
osteopenia | AUC | 0.996 | 0.942 | 0.952 | 0.996 | 0.929 | 0.940 |
(95%CI) | (0.994,0.999) | (0.890,0.994) | (0.932,0.971) | (0.992, 0.999) | (0.971, 0.987) | (0.919, 0.962) | |
Se | 0.940 | 0.833 | 0.716 | 0.893 | 0.767 | 0.638 | |
Sp | 0.984 | 0.886 | 0.960 | 0.995 | 0.957 | 0.954 | |
PPV | 0.959 | 0.758 | 0.874 | 0.985 | 0.885 | 0.841 | |
NPV | 0.977 | 0.242 | 0.898 | 0.959 | 0.905 | 0.873 | |
Ac | 0.972 | 0.870 | 0.892 | 0.967 | 0.900 | 0.866 | |
osteoporosis | AUC | 0.999 | 0.989 | 0.980 | 1.000 | 0.981 | 0.978 |
(95%CI) | (0.997,1.000) | (0.972,1.000) | (0.967,0.994) | (0.999, 1.000) | (0.957, 1.000) | (0.965, 0.990) | |
Se | 0.957 | 0.700 | 0.941 | 0.971 | 0.700 | 0.843 | |
Sp | 0.993 | 0.989 | 0.948 | 1.000 | 0.989 | 0.959 | |
PPV | 0.957 | 0.875 | 0.716 | 1.000 | 0.875 | 0.741 | |
NPV | 0.993 | 0.967 | 0.991 | 0.996 | 0.967 | 0.978 | |
Ac | 0.989 | 0.960 | 0.947 | 0.996 | 0.960 | 0.945 |