Table 1 Summary of skin cancer detection using state-of-the-art models.
From: Minimal sourced and lightweight federated transfer learning models for skin cancer detection
References | Dataset name | Dataset count | Model used | Accuracy |
|---|---|---|---|---|
ISIC 2017 | 2375 | “Graph cut algorithm” with “Naïve Bayes” classifier | 94.3% with benign cases, 91.2% with melanoma and 92.9% with keratosis | |
HAM10000 | 1619 | “VGG16”, “Inception”, “Xception”, “MobileNet”, “ResNet50” and “DenseNet161” | 97% | |
“Wuhan Union Hospital” | 6144 | Fine-tuned “ResNet152” and “InceptionResNet-V2” models | 87.42% | |
HAM10000 | 1619 | Proposed model using “MobileNet V2” and “Long Short Term Memory (LSTM)” | 85.34% | |
DermNet | 174 | Proposed CNN model | 98.6% | |
DermNet and ISIC | 23,000 | Different deep learning models | 93% | |
ISIC 2018 | 760 | Novel multimodal transformer | 92% | |
DermNet | 725 | Proposed MobileNet | 94.76% |
