Skip to main content
. 2023 Jul 5;6:119. doi: 10.1038/s41746-023-00866-z

Table 2.

Performance comparisons of the proposed EMV-3D-CNN model and the six radiologists on the validation dataset for Tasks 1 and 2.

Evaluation Index Task 1 Task 2
Model D1 D2 D3 D4 D5 D6 Model D1 D2 D3 D4 D5 D6
AUC (%) 91.3 93.9 85.1 88.4 78.6 78.9 76.5 92.9 90.4 87.1 86.6 79.9 80.1 72.7
Accuracy (%) 90.3 89.9 86.1 86.5 78.1 77.6 75.5 89.0 87.7 80.5 82.5 77.9 73.4 63.0
Sensitivity (%) 92.8 92.2 93.4 89.8 82.6 88.6 85.0 85.9 89.7 84.6 83.3 78.2 87.2 91.0
Specificity (%) 84.3 84.3 68.6 78.6 67.1 51.4 52.9 92.1 85.5 76.3 81.6 77.6 59.2 34.2
PPV (%) 93.4 93.3 87.6 90.9 85.7 81.3 81.1 91.8 86.4 78.6 82.3 78.2 68.7 58.7
NPV (%) 83.1 81.9 81.4 76.4 61.8 65.5 59.7 86.4 89.0 82.9 82.7 77.6 81.8 78.7
F1 (%) 93.1 92.8 90.4 90.4 84.1 84.8 83.0 88.7 88.1 81.5 82.8 78.2 76.8 71.4

PPV positive predictive value, NPV negative predictive value, D Doctor.