Estimation of Skin Cancer with Integrated Extended Convolutional and Recurrent Neural Network Techniques on Image Dataset

Abstract

Skin cancer is the most common and possibly fatal type of cancer that necessitates early detection through the deep-learning method. Machine learning approaches such as random forest and Naive Bayes are used to identify skin cancer. Numerous studies comparing the efficacy of Artificial Intelligence (AI) based models for automated skin cancer classification to that of human experts have laid the groundwork for the effective deployment of AI-based tools into clinical pathological practice. The detection of skin cancer using Naive Bayesian display an accuracy of 86% and the random forest method exhibits an accuracy of 87%. To improve the accuracy, an automatic skin cancer detection using an Extended Convolutional Neural Network (ECNN) technique is proposed with 12 nested processing layers, which enhances skin cancer diagnostic and detection accuracy. The ECNN network and extended Recurrent Neural Network (ERNN) display an accuracy of 94.02% and 87.32%, respectively. These investigations assess the clinical relevance of three important aspects of the existing research on melanoma reader studies: test set characteristics (composition, and out-of-distribution dataset), experimental or clinical data (metadata), and clinical symbolism of the participants. These aspects are tested set characteristics (composition, and out-of-distribution dataset); experimental or clinical data (metadata), and clinical symbolism of the participants. The search included digital biomarkers, histology, whole slide imaging, deep learning, melanoma detection, and skin cancer categorization. The results suggest that ECNN and ERNN models are more resilient and dependable when compared to existing transfer learning models.