Fig 1.
Methodology of the proposed skin cancer detection framework.
Table 1.
Summary of recent studies in skin cancer diagnostics using deep learning techniques.
Table 2.
Class distribution in HAM10000 dataset.
Table 3.
Hyperparameters utilized for data augmentation.
Table 4.
Class distribution of the training dataset before and after the application of data augmentation.
Fig 2.
Fusion model 1 (FM1) structure.
Fig 3.
Fusion model 2 (FM2) structure.
Fig 4.
Fusion model 3 (FM3) structure.
Fig 5.
Fusion model 4 (FM4) structure.
Fig 6.
Fusion model 5 (FM5) structure.
Fig 7.
Fusion model 6 (FM6) structure.
Fig 8.
Confusion matrices for the six suggested feature fusion ideas (FM1–FM6 models) on the HAM10000 dataset.
Fig 9.
Class-based ROC curves for the six suggested feature fusion ideas (FM1–FM6 models) on the HAM10000 dataset.
Fig 10.
Average ROC curves for the six suggested feature fusion ideas (FM1–FM6 models) on the HAM10000 dataset.
Table 5.
Comparison of classification metrics across different models on testing dataset.
Table 6.
Comparison of the suggested fusion models with the existing approaches in the literature.
Fig 11.
GradCAM results.
Fig 12.
SHAP values visualization.
Fig 13.
Diagnostic error analysis of the FM6 model comparing false positives and false negatives.