Table 2 Prediction performance of HRD across three models using five-fold cross-validation on the training set
Model | CT Type | AUC | Accuracy | Sensitivity | Specificity |
|---|---|---|---|---|---|
HCR | 40 kVp | 0.79 [0.74, 0.85] | 73% [68%, 78%] | 69% [54%, 84%] | 77% [60%, 94%] |
80 kVp | 0.77 [0.68, 0.87] | 71% [63%, 80%] | 68% [44%, 91%] | 74% [51%, 97%] | |
Combined | 0.79 [0.72, 0.85] | 74% [68%, 80%] | 76% [59%, 92%] | 71% [54%, 87%] | |
sDL | 40 kVp | 0.78 [0.71, 0.85] | 73% [69%, 77%] | 72% [55%, 90%] | 74% [60%, 88%] |
80 kVp | 0.71 [0.65, 0.77] | 68% [64%, 72%] | 71% [54%, 89%] | 64% [45%, 84%] | |
Combined | 0.74 [0.62, 0.86] | 71% [62%, 79%] | 76% [49%, 99%] | 66% [36%, 95%] | |
FM | 40 kVp | 0.89 [0.84, 0.94] | 84% [81%, 87%] | 83% [73%, 93%] | 86% [80%, 91%] |
80 kVp | 0.89 [0.85, 0.94] | 84% [82%, 86%] | 84% [77%, 92%] | 85% [78%, 92%] | |
Combined | 0.90 [0.84, 0.96] | 85% [82%, 88%] | 89% [81%, 97%] | 81% [75%, 87%] |
