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

Source abstract

Lung cancer is a malignant lung tumour that is characterized by the regulated growth of cells in the lung tissue. Different image processing and soft computing methods have been used for identifying lung cancer cells from medical images. Between 2018 and 2025, lung cancer diagnosis using AI and soft-computing techniques (ANN, CNN, ANFIS, fuzzy logic, metaheuristics, and hybrid systems) has progressed significantly. This PRISMA-based review systematically synthesizes recent studies to identify trends, gaps, and opportunities for developing a modified multi-agent diagnostic model. A systematic search was conducted across PubMed, IEEE Xplore, Scopus, and Google Scholar (2018–2025). Studies were included if they applied soft-computing or hybrid AI models to lung cancer detection, segmentation, or classification tasks. PRISMA 2020 guidelines were followed for screening, eligibility, and inclusion. Out of 520 initially identified records, 390 unique papers were screened; 68 underwent full-text assessment, and 32 studies met the inclusion criteria. Studies were grouped into four categories: (1) Deep CNN architectures, (2) Soft computing & fuzzy logic models, (3) Hybrid metaheuristic optimizations, and (4) Explainable AI systems. Reported accuracy ranged from 86%–99%, with recent studies (2024–2025) emphasizing explain ability and computational efficiency. The literature shows high diagnostic promise for hybrid and soft-computing approaches, though external validation and explain ability remain limited. These findings justify a modular, multi-agent framework integrating deep learning, fuzzy reasoning, and optimization agents for robust and interpretable lung cancer diagnosis.