Table 3.
A comparative analysis of skin cancer detection using ANN-based approaches.
| Ref | Skin Cancer Diagnoses |
Classifier and Training Algorithm |
Dataset | Description | Results (%) |
|---|---|---|---|---|---|
| [23] | Melanoma | ANN with backpropagation algorithm | 31 dermoscopic images | ABCD parameters for feature extraction, | Accuracy (96.9) |
| [20] | Melanoma/Non- melanoma | ANN with backpropagation algorithm | 90 dermoscopic images | maximum entropy for thresholding, and gray- level co-occurrence matrix for features extraction | Accuracy (86.66) |
| [19] | Cancerous/non- cancerous | ANN with backpropagation algorithm | 31 dermoscopic images | 2D-wavelet transform for feature extraction and thresholding for segmentation | Nil |
| [24] | Malignant /benign |
Feed-forward ANN with the backpropagation training algorithm | 326 lesion images |
Color and shape characteristics of the tumor were used as discriminant features for classification | Accuracy (80) |
| [25] | Malignant/non-Malignant | Backpropagation neural network as NN classifier | 448 mixed-type images | ROI and SRM for segmentation | Accuracy (70.4) |
| [21] | Cancerous/noncancerous | ANN with backpropagation algorithm | 30 cancerous/noncancerous images | RGB color features and GLCM techniques for feature extraction | Accuracy (86.66) |
| [18] | Common mole/non-common mole/melanoma | Feed-forward BPNN | 200 dermoscopic images | Features extracted according to ABCD rule | Accuracy (97.51) |
| [26] | Cancerous/noncancerous | Artificial neural network with backpropagation algorithm | 50 dermoscopic images | GLCM technique for feature extraction | Accuracy (88) |
| [27] | BCC/non-BCC | ANN | 180 skin lesion images | Histogram equalization for contrast enhancement | Reliability (93.33) |
| [14] | Melanoma/Non-melanoma | ANN with Levenberg–Marquardt (LM), resilient backpropagation (RBP), and scaled conjugate gradient (GCG) learning algorithms | 135 lesion images |
Combination of multiple classifiers to avoid the misclassification | Accuracy (SCG:91.9, LM: 95.1, RBP:88.1) |
| [13] | Malignant/benign | ANN meta-ensemble model consisting of BPN and fuzzy neural network | Caucasian race and xanthous-race datasets | Self-generating neural network was used for lesion extraction |
Accuracy (94.17) Sensitivity (95), specificity (93.75) |
ANN = Artificial neural network, NN = Neural network. ROI = Region of interest, SRM = Statistical region merging, GLCM = Gray level co-occurrence matrix, BPNN = Backpropagation neural network.