Hybrid Deep Learning and Feature Engineering for Early Detection of Melanoma Skin Cancer Using Multimodal Imaging Data

Skin cancer, particularly melanoma, is a severe public health concern that requires improved early diagnosis to enhance treatment results. Existing systems depend on manual assessment and limited imaging, which leads to insufficient results. To address these problems, the study provides an innovative strategy for early melanoma diagnosis utilizing multimodal imaging data that leverages Hybrid Deep Learning and Feature Engineering. The hybrid approach combines automated deep learning features for pattern extraction with autoencoder features, and their fusion is provided as SVM input, yielding exact melanoma diagnosis output. The integrative method intends to overcome system constraints by increasing sensitivity and specificity in identifying early melanoma features. The experimental results reveal that the proposed system outperforms existing systems in terms of accuracy (94.5%), precision (93.7%), recall (93.4%), and F1 (94.2%). These results highlight the system's improved diagnostic accuracy and its potential for prompt and reliable melanoma detection, which addresses the limitations of existing systems.