Table 3.
Diagnostic performances to classify tumor shrinkage pattern in testing dataset of different models based on the type of features using the Multilayer Perception (MLP) neural network.
| Model | AUC (95%CI) | Accuracy (95%CI) | Sensitivity (95%CI) | Specificity (95%CI) |
|---|---|---|---|---|
| ModelT1−DCE | 0.712 (0.644–0.771) | 0.644 (0.573–0.708) | 0.489 (0.418–0.558) | 0.783 (0.719–0.835) |
| ModelT2WI | 0.661 (0.591–0.724) | 0.606 (0.534–0.672) | 0.562 (0.491–0.630) | 0.645 (0.575–0.709) |
| ModelADCmap | 0.795 (0.732–0.846) | 0.709 (0.641–0.768) | 0.699 (0.630–0.759) | 0.718 (0.650–0.777) |
| ModelClinical | 0.611 (0.540–0.677) | 0.561 (0.489–0.629) | 0.653 (0.582–0.716) | 0.480 (0.410–0.550) |
| ModelRadiomics | 0.900 (0.849–0.935) | 0.828 (0.767–0.874) | 0.788 (0.724–0.839) | 0.864 (0.807–0.905) |
| ModelT1−DCE+Clinical | 0.743 (0.676–0.799) | 0.687 (0.618–0.748) | 0.713 (0.644–0.771) | 0.665 (0.595–0.727) |
| ModelT2WI+Clinical | 0.708 (0.640–0.768) | 0.649 (0.579–0.713) | 0.627 (0.556–0.692) | 0.670 (0.599–0.732) |
| ModelADCmap+Clinical | 0.809 (0.746–0.858) | 0.729 (0.662–0.787) | 0.699 (0.630–0.759) | 0.757 (0.690–0.811) |
| ModelRadiomics+Clinical | 0.939 (0.896–0.965) | 0.870 (0.815–0.910) | 0.840 (0.781–0.885) | 0.897 (0.846–0.932) |