Table 3 Areas under precision-recall curves of different models for occult N1 and N2 prediction
From: PET/CT based cross-modal deep learning signature to predict occult nodal metastasis in lung cancer
Models | Occult N1 prediction | Occult N2 prediction | |||||
---|---|---|---|---|---|---|---|
Validation set | External cohort | Prospective cohort | Validation set | External cohort | Prospective cohort | Biopsy cohort | |
DLNMS | 0.882 | 0.853 | 0.871 | 0.876 | 0.849 | 0.863 | 0.857 |
PET model | 0.756 | 0.731 | 0.748 | 0.753 | 0.710 | 0.741 | 0.752 |
CT model | 0.779 | 0.751 | 0.764 | 0.765 | 0.746 | 0.755 | 0.761 |
Clinical model | 0.656 | 0.612 | 0.627 | 0.694 | 0.635 | 0.648 | 0.695 |
Senior physicians | 0.504 | 0.562 | 0.538 | 0.563 | 0.583 | 0.569 | 0.550 |
Junior physicians | 0.582 | 0.590 | 0.599 | 0.514 | 0.555 | 0.534 | 0.514 |