Table 2.

Comparing skin lesion management prediction results obtained using ${\mathrm{MGMT}}_{\mathrm{infr},\mathrm{all}}$ and ${\mathrm{MGMT}}_{\mathrm{pred},\mathrm{all}}$. All the prediction models have been trained using all the input data modalities (i.e., clinical image, dermoscopic image, and patient metadata). Mean ± standard deviation reported for all the metrics for the 3-fold cross validation.

Management labels	${\mathrm{MGMT}}_{\mathrm{infr},\mathrm{all}}$					${\mathrm{MGMT}}_{\mathrm{pred},\mathrm{all}}$
	Sensitivity	Specificity	Precision	AUROC	Overall accuracy	Sensitivity	Specificity	Precision	AUROC	Overall Accuracy
NONE	0.2	0.9718	0.4444	0.8039	–	0.5	0.9831	0.7692	0.9159	–
CLNC	0.0	1.0	0.0	0.7668	–	0.7143	0.7456	0.5263	0.8090	–
EXC	0.9835	0.0921	0.634	0.7515	–	0.7243	0.7303	0.8111	0.8079	–
Average	0.3945	0.6880	0.3595	0.7741	0.6253	0.6462	0.8196	0.7022	0.8443	0.6987
3-Fold cross validation	0.3943 ± 0.0025	0.6871 ± 0.0018	0.3829 ± 0.0172	0.7758 ± 0.0178	0.6139 ± 0.0198	0.6033 ± 0.0369	0.8107 ± 0.0093	0.7104 ± 0.0184	0.8266 ± 0.0126	0.6974 ± 0.0018