Detection of Skin Lesion Disease Using Deep Learning Algorithm

Skin lesions are a part of the skin that has an unusual development or appearance contrasted with the skin around it. They may be something you are born with or something you acquire over your lifetime. They can be classified into two types: benign (non-cancerous) or malignant (cancerous). Some studies have been conducted on the computerised detection of malignancy in images. However, due to various problematic aspects such as reflections of light from the skin’s surface, the difference of colour lighting, and varying forms and sizes of the lesions, analysing these images is extremely difficult. As a result, evidence-based automatic skin cancer detection can help pathologists improve their accuracy and competency in the primitive stages of ailment. Our proposed method is to detect the early onset of skin lesions using python as a tool to detect benign (non-cancerous) or severe (cancerous) lesions using a machine learning approach. The dataset consists of nine different classes of skin lesion diseases: Melanoma (MEL), Melanocytic nevus (NV), Basal cell carcinoma (BCC), Actinic keratosis (AK), Benign keratosis (BKL), Dermatofibroma (DF), Vascular lesion (VASC), Squamous cell carcinoma (SCC), None of the above (UNK). In our proposed work, a DCNN model is created for classifying cancerous and non-cancerous skin lesions. We use techniques such as filtering, feature extraction for better categorization which will enhance the final analysis value. From our proposed model we have achieved a training accuracy of 90.7%.