Evaluating an Artificial Intelligence Model for Distinguishing Melanoma from Skin Carcinomas Using Dermoscopic Images

This study evaluates an artificial intelligence image-classification model designed to distinguish melanoma from skin carcinomas using dermoscopic images. The model was trained using 724 images, consisting of 362 melanoma images and 362 carcinoma images, and evaluated on a separate test set of 500 images, with 250 images from each category. The model correctly classified 486 of the 500 test images, achieving an overall accuracy of 97.2%. Melanoma sensitivity was 96.8%, while carcinoma sensitivity was 97.6%. These results indicate that the model performed well on the selected binary image-classification task. However, the findings are limited to the dataset and testing conditions used in this experiment. Further testing with larger, more diverse datasets and independent patient-level validation would be necessary to assess its broader reliability and potential clinical relevance. Keywords: Artificial intelligence, machine learning, skin cancer classification, melanoma, carcinoma, dermoscopic images, medical image analysis, image classification Preprint Notice: This manuscript is an open-access scientific preprint published in the Global Youth Science Journal (GYSJ) Preprints repository. It has not undergone formal journal peer review. Permanent Repository URL: https://globalyouthsciencejournal.app/preprint/202611.0013