The Smart Dermatoscope: A Novel AI-Enhanced System for Early Skin Cancer Detection

This study investigates the application and fine tuning of transfer learning with EfficientNet architecture in the field of skin cancer detection .In order to ensure the accuracy and efficiency of this system we used the pre-trained model EfficientNet-B0.It is based on CNN architecture which was later fine tuned on the ISIC 2024 dataset.Data augmentation is applied to overcome the scarcity of true positive cases in the data set , it was applied using the Albumentations python based image augmentation library which also makes the model generalise better. Stratified K-fold cross validation (K=5) was used to ensure robust performance evaluation.The model achieved a mean Area Under The ROC Curve (AUC) of 0.88 across the cross validation fold. These results show that the model EfficinetNet-B0, with the assistance of transfer learning and data augmentation can be used as an effective tool to assist dermatologists to identify skin cancer.