Evaluation of Modified VGG16 Learning Model for Classifying Skin Cancer Lesions

One of the most serious types of cancer that is prevalent around the world is skin cancer. It is widely known that skin cancer characterized by uncontrolled cell and tissue growth, presenting a challenge in detecting or diagnosis skin cancer in early stage. Accurate segmentation of skin lesions in dermatoscopic images holds great importance in diagnosis and assessment of skin cancer. Thus, Fast, precise, and real-time algorithms are essential in effectively detecting skin cancer to aid physicians in their treatment decision-making process. In this paper, we evaluated the performance of the modified VGG16 with using three common convolutional neural networks(CNN) in detecting the skin cancer with dermoscopic skin lesions images. In this study, the Kaggle dataset is used to classify skin cancer images. To classify skin lesions, the model modified VGG16 is compared to the most common models such as VGG16, mobileNetV2 and Inception ResNet v2. All models are trained on the same dataset and their performance is evaluated using metrics such as: accuracy, precision, F1-score, and recall. Finally, The comparative analysis revealed a promising performance of the modified VGG16 model in detecting multiclass skin cancer, which highlights its promising potential in classifying.