<![CDATA[Lung cancer remains the leading factors that incur cancer-related deaths worldwide, mainly due to late-stage detection. In 2022, lung cancer affected almost 2.5 million people, with mortality of more than 1.8 million. However, existing prognostic methods are typically invasive, costly, and time-consuming, hindering effective early detection. Therefore, this research proposes a non-invasive prognostic approach using salivary biomarkers to detect lung cancer via Graph Convolutional Networks (GCNs). By transforming features into graph node representations, the proposed algorithm can model feature dependencies and topological relationships, enabling more effective pattern recognition than conventional classifiers. The proposed algorithm also applies feature selection to reduce computational complexity. The evaluation results show that the proposed algorithm achieves 95.65% accuracy, a macro F1-score of 95.62%, and a Matthews Correlation Coefficient of 0.9434. A comparative analysis shows that the proposed algorithm outperforms other graph-based architectures in terms of classification performance and computational complexity.]]>
Copyrights © 2026
