Ensemble Deep Learning Framework for Classification of Skin Lesions

Skin cancer is one of the majority prevalent malignancies that can be diagnosed visually and then further investigated with dermoscopic analysis and other procedures. Several skin-lesion (SL) classification methods using deep learning (DL) based on convolution-neural-network (CNN) and annotated skin photos illustrate modified results because visual observation provides the opportunity to use artificial-intelligence (AI) to intercept the various skin images at an early stage. In this regard, the research proposes a trustworthy method for identifying skin cancer using dermoscopy images, with the goal of bettering the visual perception and diagnostic abilities of medical experts. SL region of interest (RoI) segmentation from dermoscopy pictures was performed using swarm intelligence (SI) algorithms, and the RoI marked as the best segmentation oucome was subjected to feature extraction using the Grasshopper Optimization Algorithm (GOA). Using CNN and the ISIC-2017, ISIC-2018, and PH-2 data sets, the SLs are divided into two categories. Classification accuracy-sensitivity,-specificity-F -measure, precision, MCC, dice coefficient, and Jaccard index are used to evaluate the results of the proposed segmentation and classification strategies, with an average of 98.42%, 977.33%, and 0.97040%, respectively. Our proposed model outperforms the state-of the-art in every metric we considered