Table 1 The state-of-the-art techniques for detection skin cancer.
From: Enhanced melanoma and non-melanoma skin cancer classification using a hybrid LSTM-CNN model
Refs. | Dataset | Model | Activation function | Accuracy | Classes of skin lesion |
|---|---|---|---|---|---|
ISIC | CNN + SVM | ReLu | 91% | Benign, malignant | |
PH2, DermIS | CNN | Sigmoid | 94.9% | Melanoma, non-melanoma | |
SIC-2019 ISIC-2020 | DCNN | ReLu | – | Melanoma, non-melanoma | |
7-Point, Med-node, PAD-UFES-20 and PH2 | ANN | Â | 96.7% | Melanoma, nevus, dysplastic nevus | |
DermIS | LSTM | – | 99.4% | Melanoma, benign lesions | |
ISIC 2017 | CNN + LSTM | Sigmoid | 94.6% | Melanoma, benign lesions | |
18274 dermoscopy images | CNNs and LSTMs | Sigmoid | 93.41% | Benign, malignant | |
HAM 10000 | CNN + RNN |  | 94% | Melanoma, nevi, dermatofibroma, seborrheic keratosis, BCC, and SBC | |
ISIC | CNN + SVM Sparse Coding, |  | 93.1% | Melanoma, atypical nevi, benign lesions | |
ISIC 2019 | DCNN + RF + Naïve Bayesian | Sigmoid | 99.5% | Melanoma, solar lentigo, seborrheic keratosis |
