Lung Cancer Prediction Through Feature Extraction And Classification Using Enhanced Graph Convolutional Networks

Abstract

Lung cancer remains a significant global health concern, necessitating the development of advanced detection and classification methods. This study proposes a comprehensive approach combining Hyper Correlation Feature Selection (HCFS) for feature extraction and Enhanced Graph Convolutional Networks (E-GCN) for classification. HCFS is employed to identify the most relevant features from high-dimensional lung cancer datasets by leveraging inherent correlations among features. These selected features are then utilized with E-GCN, a state-of-the-art deep learning architecture capable of effectively modeling complex relationships in data represented as graphs. We applied this approach to a diverse dataset comprising clinical and imaging features associated with lung cancer patients. Through rigorous experimentation and evaluation, our results demonstrate the efficacy of HCFS in extracting informative features that significantly contribute to accurate lung cancer classification. Additionally, integrating these features with E-GCN enables the development of a robust classification model with enhanced performance. Our study highlights the potential of combining feature selection techniques with advanced deep learning architectures for improved lung cancer diagnosis and prognosis prediction.