Application of Deep Learning Methodology for Early Detection of Melanoma

Early diagnosis of melanoma warning signs can help prevent cancer from spreading to other parts of the body. Dermatologists use their skills and knowledge to assess skin lesion imaging data to identify malignant melanoma. The error can be overcome by using computer vision and deep learning algorithms for automatic segmentation and detection for accurately identifying the disease. Deep learning in melanoma detection can speed up diagnosis, buy time for medical treatment, lowering treatment costs, allow for home-based diagnostics and monitoring, and provide the best value to the patient. Medical image segmentation or detection based on deep learning methods is an efficient method that outperforms human-level accuracy. Deep learning algorithms are used in dermoscopy for the improvement of analysis, segmentation, and melanoma classification. These algorithms are trained using deep neural networks on many annotated malignant and benign melanoma images. This study presents a comprehensive and systematic literature review of the conventional approach used in melanoma detection using CNN. A large number of datasets are available to detect melanoma and some of them are ISIC archive, HAM10000, PH2, MED-NODE, and other image libraries are widely used datasets. Dice-coefficient, sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC-ROC) are the commonly adopted metrics by different researchers across the globe