FMICS: A Robust Learning Methodology to Predict Skin Cancer Disease Using Fused Melanoma Image Classification Scheme

Skin cancer (SC), particularly melanoma, is a significant health issue due to its high mortality rate when not detected early. This study introduces a novel approach, FMICS: A Robust Learning Methodology (Hybrid ANN + XGBoost) for predicting SC through fused melanoma image classification. The methodology integrates Artificial Neural Networks (ANN) for feature extraction and XGBoost for classification, leveraging the strengths of both models. Preprocessing techniques such as image filtering, contrast enhancement, and hair removal were applied to improve image quality. The study utilized dermoscopic images from the PH2 and HAM10000 datasets, containing both benign and malignant lesions. The hybrid model achieved superior performance compared to individual models, with an accuracy of 94.80%, sensitivity of 94.90%, and specificity of 94.70%. Data augmentation techniques, including rotation, scaling, and flipping, were applied to enhance the model's robustness. Evaluation metrics such as the F1-score and ROC-AUC were used to assess the performance. The hybrid approach significantly reduced false negatives, making it highly reliable for melanoma detection. These results suggest that the FMICS methodology offers a promising solution for accurate and efficient SC diagnosis, with potential applications in clinical and mobile environments.