Abstract
More and more chromosomal and metabolic abnormalities are now known to cause cancer, which is typically fatal. Anybody component may become infected
by tumour cells, which can be fatal. One of the most prevalent types of cancer is skin cancer, and its prevalence is rising around the globe. Early diagnosis
and delineation of the lesion margins are crucial for precise malignant region identification and clinical treatment of skin lesions. Skin cancer incidence is
greater than average, particularly melanoma, which is more dangerous because of its high rate of metastasis. Therefore, early detection is essential for treating it before malignancy develops. The analysis and segmentation of lesion boundaries from dermoscopic images is done in order to solve this issue. A variety of techniques have been utilised, from textural assessment of the photographs to visual assessment of the images. However, due to the sensitivity involved in surgical interventions or drug distribution, the accuracy of these techniques is poor for real clinical therapy. This offers a chance to create an automatic system that is accurate enough to be applied in a clinical environment. Epithelial tissue and basal cell carcinomas, as well as melanoma, which is medically severe and causes the majority of deaths, are the main subtypes of skin cancer. Monitoring for skin cancer is therefore essential. Machine learning is one of the greatest ways to quickly and precisely identify skin cancer. To use the ISIC2018 database, the convolution neural network (CNN) deep learning technique was employed in this study to identify the two main categories of tumours, malignant and benign. Skin lesions, comprising benign and malignant tumours, are included in this database. The images were initially enhanced and edited using ESRGAN. The pre- processing stage involved resizing, normalising, and augmenting the images. Using a CNN approach, skin lesion images might be categorised based on an accumulation of data collected after numerous repetitions. The experimental results show that the proposed methodology performance is better than existing methodologies.
Keywords
NA