RLMD-PA: A Reinforcement Learning-Based Myocarditis Diagnosis Combined with a Population-Based Algorithm for Pretraining Weights

Table 2

5-CV classification performances (accuracy, recall, and precision) obtained for automated myocarditis detection using various combinations of deep learning models with the Z-Alizadeh Sani myocarditis dataset.

	Accuracy					Recall					Precision
Method	Min	Median	Max	Mean	Std.dev.	Min	Median	Max	Mean	Std.dev.	Min	Median	Max	Mean	Std.dev.
CNN-KCL [3]	0.783	0.811	0.846	0.810	0.024	0.732	0.738	0.807	0.751	0.032	0.704	0.752	0.789	0.745	0.032
CNN + random weight	0.755	0.770	0.807	0.772	0.021	0.695	0.713	0.755	0.717	0.213	0.666	0.685	0.737	0.691	0.029
CNN + ABC	0.799	0.803	0.845	0.815	0.020	0.741	0.766	0.814	0.771	0.027	0.726	0.729	0.783	0.746	0.027
CNN + RL	0.821	0.829	0.869	0.840	0.021	0.762	0.798	0.835	0.801	0.028	0.745	0.772	0.819	0.779	0.029
RLMD-PA (CNN + ABC + RL)	0.862	0.884	0.912	0.886	0.020	0.837	0.869	0.879	0.863	0.017	0.804	0.837	0.886	0.840	0.034