Automated Skin Lesion Analysis and Classification Using AI

Skin lesions are becoming a major global health concern, and survival rates can be greatly increased by employing dermoscopic pictures for early identification. It is a difficult undertaking to classify these skin lesions from dermoscopic images. The application of deep learning models has significantly improved this classification's accuracy. Because convolutional neural networks (CNNs) are better at extracting features, which improves classification results, they have been shown to be quite useful in this field. In light of this, the article attempts to create deep learning (DL) algorithms using artificial intelligence (AI) to differentiate between skin lesions as benign or malignant in dermoscopic pictures. The study presents OSSAE-BPNN, an automated method for the detection and classification of skin lesions. This method uses a backpropagation neural network (BPNN) for classification and an optimized stacked sparse autoencoder (OSSAE) for feature extraction. To determine the damaged lesion region, the multi-level thresholding segmentation methodology is part of the proposed OSSAE-BPNN method. The OSSAE performs the feature extraction, and the BPNN handles the categorization. Furthermore, the SGO (seagull optimization) technique is used to optimize the SSAE model's parameters. Extensive testing were conducted on a standard dataset to show the OSSAE-BPNN method's efficacy. The OSSAE-BPNN methodology fared better than other current methods across a range of evaluation parameters, according to the results.