A Computer-Aided Diagnosis System Using Deep Learning for Multiclass Skin Lesion Classification

Table 1

Classification accuracy of fine-tuned ResNet-50 deep features using augmented HAM10000 dataset.

Classifier	Recall rate (%)	Precision rate (%)	FNR (%)	AUC	Accuracy (%)	Time (sec)	F1-score (%)
LSVM	86.42	86.85	13.57	0.988	86.5	742.5	86.63
QSVM	92.00	92.14	8.00	0.992	91.7	1046.1	92.07
CSVM	93.14	93.14	6.858	0.994	92.7	1190.3	93.14
MGSVM	89.57	90.00	10.42	0.988	89.3	1906.8	89.78
CKNN	53.25	65.28	46.75	0.898	60.8	274.5	58.65
CKNN	80.42	79.00	19.57	0.967	78.7	287.6	79.70
WKNN	85.14	84.42	14.85	0.98	83.6	262.3	84.78
ESKNN	93.14	92.57	6.858	0.99	92.3	4514.7	92.85
EBT	55.28	86.71	44.71	0.974	57.1	1546.0	86.49
ESD	86.28	57.00	13.71	0.85	86.1	839.8	56.12

The bold value represents best ones.