Utilizing dermoscopy methods with machine learning and deep learning algorithms to diagnose skin lesions

Abstract Artificial intelligence is fast-growing and applied in a wide range of industries nowadays, including the healthcare sector. Dermatology is one of the areas where AI has a big influence, particularly when it comes to dermoscopy-based skin lesion diagnosis. This paper aims to develop useful techniques for disease image classification that make use of deep learning and machine learning techniques. Continuously, looking at and making suggestions for improvements is to raise the model’s efficacy during the training phases. The convolutional Neural Network (CNN) model was developed, trained, and evaluated on the Skin Cancer MNIST: HAM10000 data set, and subsequently deployed on Kaggel software using Python and the Keras framework. The results are mostly 0.95 or higher, models with Precision and Recall measurements are relatively stable. In particular, for cases like actinic keratosis, benign skin tumors, and vascular lesions, the values of Precision and Recall both reached level 1, indicating that there were no instances of incorrect predictions.