Table 1 Summary of key related works.
From: Boosting skin cancer diagnosis accuracy with ensemble approach
Study | Method | Approach | Results/Key findings |
|---|---|---|---|
Esteva et al. (2017)17 | SVM with curated features | Binary classification | Achieved comparable performance to deep learning models; effective in resource-constrained environments |
Arora et al. (2022)18 | Bag of Features (BoF) with SVM | Binary classification | Used BoF with SVM to classify skin lesions based on color and texture features; effective for limited data settings |
Kalpana et al. (2023)19 | SVM and RF ensemble (OESV-KRF) | Ensemble method | Demonstrated high efficiency, accuracy, and adaptability, suitable for real-time and resource-limited applications |
Mahbod et al. (2019)21 | RF with ensemble framework | Binary classification | Achieved high accuracy in melanoma versus non-melanoma classification; less computationally intensive than deep learning models |
Goyal et al. (2019)25 | Random Forest (RF) | Ensemble learning | RF ensemble showed high classification accuracy for melanoma, leveraging bagging to reduce variance |
Dhivyaa et al. (2020)22 | RF, SVM with CNNs | Hybrid model | Improved classification accuracy for melanoma versus benign lesions by combining CNNs and RF in a hybrid model |
Gamil et al. (2024)26 | AdaBoost | Boosting | Improved sensitivity and specificity in skin cancer classification by focusing on difficult cases iteratively; effective for imbalanced datasets |
Chang et al. (2022)27 | Gradient Boosting with deep features | Boosting | Achieved high accuracy and specificity in melanoma detection; effective in handling class imbalance common in medical imaging |
Shorfuzzaman et al. (2022)28 | Stacking ensemble of CNNs | Stacking | Enhanced classification accuracy by leveraging CNNs and traditional models in a stacking ensemble, capturing both high-level patterns and precise boundaries |
Bhowmik et al. (2019)29 | Max Voting ensemble with IG-ResNeXt-101, SWSL-ResNeXt-101, ECA-ResNet-101, DPN-131 | Max Voting | Achieved higher accuracy in melanoma detection than individual models; improved robustness through majority voting |
Perez et al. (2022)30 | Genetic Algorithm (GA) | Feature selection | Enhanced model generalization and reduced overfitting by selecting relevant features for skin cancer diagnosis |
Rumelhart et al. (1986)23 | Multi-Layer Perceptron (MLP) | Neural network for non-linear feature extraction | Foundational work on backpropagation for learning complex non-linear patterns, valuable for medical image analysis. |
LeCun et al. (2015)24 | Deep learning models (CNN) | Skin lesion classification | Demonstrated CNN’s efficacy in learning hierarchical features from image data, yet highlighted high data and computational requirements |
Barata et al. (2018)32 | Color and texture feature extraction | Hand-crafted feature extraction | Highlighted the importance of color and texture features |
