Skin cancer detection from dermoscopic images using Deep Siamese domain adaptation convolutional Neural Network optimized with Honey Badger Algorithm

Melanoma skin cancer poses significant challenges in terms of early detection and accurate diagnosis. It often goes unnoticed in its early stages, leading to advanced diagnoses with poorer treatment outcomes. Diagnosis is subjective and can vary among healthcare professionals, while limited accessibility to dermatologists further delays diagnosis. Addressing these challenges requires innovative solutions, leveraging technology and artificial intelligence, to improve early detection, enhance accuracy, and ultimately improve patient outcomes in melanoma skin cancer. In this paper, Skin cancer identification from dermoscopic images utilizing Deep Siamese domain adaptation convolutional Neural Network optimized with Honey Badger Algorithm is proposed. The proposed method initially performs input image pre-processing to remove label noise and lighting problems. The segmentation was then given the output of the pre-processing. To separate the ROI region, hesitant fuzzy linguistic bi-objective clustering is employed. The improved non-subsampled Shearlet transforms (INSST) region to extract features using the segmented ROI region. Skin cancer and healthy skin are distinguished using the Deep Siamese domain adaptation convolutional Neural Network (DSDACNN). The Honey Badger Algorithm is used to optimize the weight parameters of the DSDACNN. The proposed SKD-DSDACNN-HBA method is carry out in Python. The performance of the proposed SKD-DSDACNN-HBA method attains 15.64%, 20.07% and 25.5% higher F1-score, 24.35%, 29.33% and 35.29% higher Computational time, 24.72%, 29.32% and 36.66% higher AUC and 33.55%, 28.52% and 19.85% lower Error rate while compared with existing methods, like fuzzy k means clustering (SKD-FKMC), handcrafted and non-handcrafted features (SKD-HNHF) and deep learning features with improved moth flame optimization (SKD-DLF-IMFO).