Table 2.

DenseNet161 model validation performance from 2-class hyperparameter tuning experiments following fourfold cross-validation and 5 independent repetitions.

Validation Rank	Hyperparameters			Validation Performance ($\mu \pm 95CI$)
	Input Shape	Num. Epochs	Batch Size	AUC	Accuracy	F1 Score	Sensitivity/Recall	Specificity	Precision
1	128 × 128	300	32	0.913 ± 0.005	0.840 ± 0.008	0.763 ± 0.02	0.774 ± 0.020	0.874 ± 0.013	0.758 ± 0.023
2	128 × 128	300	64	0.900 ± 0.006	0.813 ± 0.007	0.750 ± 0.016	0.841 ± 0.014	0.800 ± 0.008	0.677 ± 0.015
3	256 × 256	600	64	0.894 ± 0.008	0.811 ± 0.011	0.737 ± 0.020	0.753 ± 0.022	0.854 ± 0.025	0.732 ± 0.036
4	128 × 128	300	64	0.892 ± 0.006	0.807 ± 0.008	0.732 ± 0.016	0.773 ± 0.016	0.831 ± 0.014	0.700 ± 0.025
5	256 × 256	300	32	0.888 ± 0.010	0.810 ± 0.011	0.732 ± 0.028	0.730 ± 0.023	0.868 ± 0.016	0.741 ± 0.033
6	128 × 128	100	64	0.884 ± 0.010	0.794 ± 0.011	0.732 ± 0.020	0.844 ± 0.016	0.769 ± 0.013	0.648 ± 0.021
7	128 × 128	300	128	0.870 ± 0.008	0.802 ± 0.012	0.717 ± 0.019	0.749 ± 0.017	0.827 ± 0.020	0.693 ± 0.032
8	256 × 256	300	64	0.866 ± 0.009	0.802 ± 0.009	0.700 ± 0.027	0.706 ± 0.032	0.849 ± 0.016	0.709 ± 0.030
9	256 × 256	300	64	0.866 ± 0.009	0.802 ± 0.009	0.700 ± 0.027	0.706 ± 0.032	0.849 ± 0.016	0.709 ± 0.030
10	512 × 512	600	64	0.862 ± 0.010	0.779 ± 0.023	0.682 ± 0.032	0.709 ± 0.042	0.813 ± 0.047	0.696 ± 0.057
11	256 × 256	300	128	0.830 ± 0.012	0.775 ± 0.009	0.657 ± 0.028	0.658 ± 0.035	0.834 ± 0.018	0.672 ± 0.027
12	256 × 256	300	128	0.830 ± 0.012	0.775 ± 0.009	0.657 ± 0.028	0.658 ± 0.035	0.834 ± 0.018	0.672 ± 0.027
13	512 × 512	300	64	0.824 ± 0.010	0.734 ± 0.011	0.661 ± 0.020	0.778 ± 0.017	0.712 ± 0.012	0.576 ± 0.020