Automatically Early Detection of Skin Cancer using Hybrid Neural Network Models

Abstract

The prevalence of skin cancer is increasing globally, rendering it one of the most prevalent forms of cancer in humans. This occurs when cells proliferate uncontrollably due to anomalous DNA modifications. The probability of effective treatment for skin cancer significantly rises with early detection, rendering a timely and accurate diagnosis essential. This study introduces an innovative computer-aided diagnostic system employing a four-stage methodology—preprocessing, segmentation, feature extraction, and model training—to enhance early detection. The dataset undergoes cleansing during preprocessing to ensure high-quality input, and the ACT is employed for segmentation. They employ the GLCM approach to capture significant visual characteristics of lesions. They introduce a paradigm for semantic modelling, termed CNRNN, to use the optimal characteristics of both recurrent and convolutional neural networks. This framework supports two models that improve performance in phrase matching and classification: an attention-based response selection model and a novel model. The proposed system surpasses existing models with an accuracy of 96.28 percent. These results underscore the potential of employing DL techniques for the early detection of skin cancer, offering a powerful tool to assist in clinical decision-making and improve patient outcomes through expedited and more accurate diagnoses. Keywords—Active Contour Technique (ACT), Grey Level Co-occurrence Matrix (GLCM), Natural Language Preprocessing (NLP), Computer Aided Diagnosis (CAD).