Table 4 Performance of the deep learning classification model on the internal and external test sets
From: Automatic segmentation of fat metaplasia on sacroiliac joint MRI using deep learning
Data sets | AUC | Accuracy (%) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | Cohen’s κ |
|---|---|---|---|---|---|---|---|
Internal test set | 0.876 (0.811–0.942) | 81.25 (73.44–89.06) | 88.52 (82.15–94.89) | 68.57 (59.28–77.86) | 83.08 (75.58–90.58) | 77.42 (69.06–85.78) | 0.504 (0.335–0.673) |
External test set | 0.799 (0.696–0.902) | 77.59 (66.85–88.32) | 91.89 (84.87–98.92) | 52.38 (39.53–65.23) | 77.27 (66.49–88.06) | 78.57 (68.01–89.13) | 0.477 (0.242–0.712) |