Enhancing Skin Cancer Diagnosis with Ensemble Deep Convolutional Neural Networks: A Multi-Model Optimization Approach

This study focuses about the identification and classification of skin cancer. It involves the use of datasets that contain images with labels and classes. The research utilizes pre-trained convolutional neural network (CNN) models, as well as integrated models. Skin cancer, a neoplastic disorder, is commonly observed on sun-exposed skin due to the proliferation of aberrant skin cells. The three most common types of skin cancer are melanoma, basal cell carcinoma, and squamous cell carcinoma. As per the American Cancer Society, it is projected that the United States will witness 99,780 fresh cases of melanoma in the year 2022. Basal cell carcinoma constitutes more than 80% of all cases of skin cancer, rendering it the predominant type. The study utilizes deep Convolutional Neural Network (CNN) models including VGG19, Xception, VGG16, InceptionV3, and ResNet50. These models are trained using image inputs that are resized to 224 x 224 pixels. The results of the segmentation algorithm can be evaluated using a combination of computational tools and clinical characteristics to accurately detect and categorize skin cancer.