Table 3 The levels of prediction for gender-specific model
Model | Male models | Female models | ||||||
|---|---|---|---|---|---|---|---|---|
Pre-trained models | Non-pre-trained models | Pre-trained models | Non-pre-trained models | |||||
mn40_as | mn30_as | mn4.0 | mn3.0 | mn40_as | mn30_as | mn4.0 | mn3.0 | |
AUC (area under the curve) | ||||||||
AUC average (95% CI) | 0.7550 (0.6056, 0.9045) | 0.8010 (0.6589, 0.9432) | 0.7429 (0.6262, 0.8596) | 0.6905 (0.5358, 0.8451) | 0.7622 (0.6169, 0.9075) | 0.7572 (0.6578, 0.8567) | 0.7679 (0.6426, 0.8931) | 0.7100 (0.5595, 0.8605) |
AUC max in 10 folds | 1.0000 | 1.0000 | 1.0000 | 1.0000 | 1.0000 | 0.9779 | 0.9722 | 0.9559 |
Accuracy (%) | ||||||||
Accuracy average (95% CI) | 79.44 (69.01, 89.88) | 85.13 (78.07, 92.19) | 78.61 (70.21, 87.01) | 69.96 (58.61, 81.30) | 69.17 (58.35, 79.99) | 69.16 (61.76, 76.57) | 69.16 (62.42, 75.89) | 69.30 (61.13, 77.48) |
Accuracy max in 10 folds | 100.00 | 100.00 | 96.00 | 87.50 | 93.10 | 88.00 | 78.57 | 88.00 |
mAP (mean average precision, %) | ||||||||
mAP average (95% CI) | 78.13 (65.24, 91.03) | 82.36 (70.38, 94.34) | 76.66 (66.13, 87.19) | 74.88 (62.57, 87.20) | 75.69 (63.10, 88.29) | 75.86 (66.33, 85.40) | 74.65 (64.61, 84.69) | 71.55 (59.37, 83.74) |
mAP max in 10 folds | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 97.19 | 97.49 | 95.44 |
Sensitivity (%) | ||||||||
Sensitivity average (95% CI) | 79.79 (69.85, 89.73) | 84.95 (77.73, 92.16) | 78.61 (70.21, 87.01) | 69.96 (58.61, 81.30) | 69.42 (58.74, 80.10) | 69.16 (61.76, 76.57) | 69.16 (62.42, 75.89) | 69.30 (61.13, 77.48) |
Sensitivity max in 10 folds | 100.00 | 100.00 | 96.00 | 87.50 | 93.10 | 88.00 | 78.57 | 88.00 |
Specificity (%) | ||||||||
Specificity average (95% CI) | 73.22 (59.93, 86.50) | 75.92 (62.97, 88.86) | 68.75 (57.86, 79.64) | 65.39 (54.48, 76.30) | 61.55 (49.89, 73.21) | 64.78 (56.87, 72.70) | 50.00 (50.00, 50.00) | 54.65 (46.67, 62.63) |
Specificity max in 10 folds | 100.00 | 100.00 | 91.67 | 87.50 | 92.86 | 81.25 | 50.00 | 84.56 |
Precision (%) | ||||||||
Precision average (95% CI) | 73.57 (60.88, 86.25) | 74.68 (60.26, 89.10) | 71.37 (56.10, 86.63) | 68.61 (57.10, 80.11) | 64.87 (49.88, 79.86) | 66.26 (55.97, 76.55) | 34.58 (31.21, 37.94) | 41.84 (28.80, 54.89) |
Precision max in 10 folds | 100.00 | 100.00 | 97.73 | 90.00 | 86.36 | 92.50 | 39.29 | 87.30 |
F1 Score | ||||||||
F1 Score average (95% CI) | 0.7971 (0.6997, 0.8946) | 0.8317 (0.7407, 0.9228) | 0.7744 (0.6855, 0.8632) | 0.6973 (0.5957, 0.7989) | 0.6611 (0.5449, 0.7772) | 0.6878 (0.6201, 0.7555) | 0.5689 (0.4829, 0.6548) | 0.5962 (0.4874, 0.7051) |
F1 Score max in 10 folds | 1.0000 | 1.0000 | 0.9576 | 0.8730 | 0.9202 | 0.8710 | 0.6914 | 0.8777 |
Loss | ||||||||
Loss average (95% CI) | 0.8648 (0.4610, 1.2690) | 0.5064 (0.2040, 0.8090) | 1.1312 (0.6060, 1.6560) | 1.6051 (0.8250, 2.3860) | 0.9823 (0.5800, 1.3850) | 1.2326 (0.4640, 2.0010) | 1.0512 (0.6140, 1.4890) | 0.9657 (0.5680, 1.3630) |
Loss max in 10 folds | 1.6027 | 1.1415 | 2.5325 | 4.3304 | 2.0750 | 4.0219 | 2.3448 | 1.9062 |
Train accuracy (%) | ||||||||
Train accuracy average (95% CI) | 99.94 (99.80, 100.08) | 100.00 (100.00, 100.00) | 99.97 (99.91, 100.04) | 99.97 (99.90, 100.04) | 100.00 (100.00, 100.00) | 99.92 (99.81, 100.04) | 99.92 (99.81, 100.04) | 99.81 (99.61, 100.00) |
Train accuracy max in 10 folds | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 | 100.00 |
Train loss | ||||||||
Train loss average (95% CI) | 0.0065 (− 0.0037, 0.0168) | 0.0033 (0.0018, 0.0049) | 0.0013 (− 0.0004, 0.0029) | 0.0016 (− 0.0004, 0.0036) | 0.0150 (0.0014, 0.0287) | 0.0284 (0.0045, 0.0523) | 0.0298 (0.0046, 0.0550) | 0.0357 (− 0.0047, 0.0760) |
Train loss max in 10 folds | 0.0474 | 0.0078 | 0.0076 | 0.0092 | 0.0618 | 0.0799 | 0.0998 | 0.1849 |