A Modified Multi-Agent Model for Diagnosis of Lung Cancer Using Soft Computing Techniques

Lung cancer remains the leading cause of cancer-related mortality worldwide, necessitating accurate and efficient diagnostic tools to improve patient outcomes. Existing computer-aided diagnostic models often rely on rigid mathematical approaches, which struggle to handle the complexity and uncertainty associated with minute pulmonary nodules and accurate severity-level determination in clinical images. Furthermore, common weaknesses in existing models include low explainability, high computational overhead, and poor scalability. This research aimed to address these limitations by enhancing an existing multi-agent model based on soft computing techniques (Fuzzy Logic, Artificial Neural Networks, and Genetic Algorithms) to diagnose lung cancer, determine its severity, and improve diagnostic accuracy and computational efficiency. The proposed system utilizes a sixagent workflow for image preprocessing, feature fusion, and hybrid classification of CT scans from the LIDC-IDRI dataset. The hybrid model, which integrates Convolutional Neural Network (CNN) feature extraction with a Genetic Algorithm (GA)-optimized Fuzzy Inference System (FIS), achieved a superior Area Under the Curve (AUC) score of 0.91. This performance significantly outperformed the standalone CNN-only model (AUC = 0.83) and the Fuzzy-only system (AUC = 0.76). The results validate the synergistic advantage of the hybrid framework in delivering enhanced diagnostic precision, demonstrating its potential utility as a robust clinical decision-support tool.