Table 6.

Comparison of related articles for detecting skin cancer using different types of imaging, differentiating the classification problems, where MM: Malignant melanoma; Malignant includes MM, Basal-cell carcinomas and Squamous-cell carcinomas.

Reference	Methodology	Classification problem	Dataset	Technology	Performance
[61]	Nasnet Mobile with transfer learning	MM vs. non-MM	HAM10000	Dermoscopy	97.90% accuracy
[62]	MobileNetV2-based transfer learning	MM vs. benign	ISIC2020	Dermoscopy	98.20% accuracy
[63]	CNN DenseNet-121 with multi-layer perceptron (MLP)	MM vs. non-MM	ISIC 2016, ISIC 2017, PH2 y HAM10000	Dermoscopy	Accuracy of 98.33%, 80.47%, 81.16% and 81.00% on PH2, ISIC 2016, ISIC 2017 and HAM10000 datasets.
Own implementation	ResNet152V2-based transfer learning	MM vs. benign	HAM10000	Dermoscopy	90.63% accuracy
Own implementation	ResNet152V2-based transfer learning	MM vs. benign	UdeC	Low-quality visible images	72.49% accuracy
[15]	Deep learning	MM vs.benign		Passive IRT	96.91% accuracy
[15]	Deep learning	Malignant vs.benign		Passive IRT	57.58% accuracy
Proposed method using automatic U-Net segmentation	Machine learning	Malignant vs.benign	UdeC	Active IRT	75.29% accuracy
Proposed method using manual expert segmentation	Machine learning	Malignant vs.benign	UdeC	Active IRT	86.61% accuracy