Skin Lesion Classification of Dermoscopic Imagesusing Machine Learning and Convolutional Neural Network

Abstract Detecting dangerous illnesses connected to the skin organ, particularly malignancy, requires the identification of pigmentedskin lesions. If a patient with a pigmented lesion is diagnosed as having or at risk of getting skin cancer, immediate steps canbe done to reduce their risk or eliminate the malignancy if it is detected early. Dermoscopy and dermatologist are the mostpopular methods for diagnosing skin cancer. Image detection techniques and computer classification capabilities can boostskin cancer detection accuracy. The dataset used for this research work is based on the HAM10000 dataset which consists of10015 images. We have analyzed the classification accuracy of the Machine Learning algorithms and Convolutional NeuralNetwork models. We have concluded that Convolutional Neural Network provides better accuracy compared to other machinelearning algorithms implemented in the proposed work. Precision, Recall, and F1-Score were chosen as evaluation metrics,along with Accuracy. In the proposed system, as the highest, we obtained an accuracy of 95.18% with the CNN model. Thek-fold cross-validation method is used to validate the accuracy obtained by the Machine Learning algorithm. The proposedwork helps early identification of seven classes of skin disease and can be validated and treated appropriately by medicalpractitioners.