Artificial Intelligence in Oncologic Radiology: Past, Present and Future A Brief History of AI in Radiology

Abstract

Introduction: Artificial intelligence (AI) has undergone remarkable evolution in radiology since the 1960s, shifting from fully autonomous diagnostic goals to becoming a powerful assistive tool for radiologists. This transformation, particularly through artificial neural networks and deep learning methods, has significantly enhanced image interpretation, workflow efficiency and patient outcomes in oncologic imaging. Aims and Objectives: The aim of this paper is to review the historical development, current applications and future directions of AI in oncologic radiology, with a particular focus on breast and lung cancer diagnostics. The objective is to highlight how AI enhances diagnostic accuracy, supports clinical decision-making and paves the way toward personalized radiology. Materials and Methods: The article is a narrative literature review analyzing the progression of AI technologies in radiology. It includes studies and examples of deep learning systems, computer-aided detection (CADe) and diagnosis (CADx), with references to key AI models, such as CNNs and the CANARY system and their applications in various imaging modalities (CT, MRI, PET, X-ray). Results: AI systems, particularly deep learning and CAD tools, have demonstrated high diagnostic accuracy comparable to experienced radiologists. Applications in breast cancer imaging showed increased sensitivity and specificity in tumor detection. In lung cancer, AI enabled early detection, differentiation of benign vs malignant nodules and quantification of tumor heterogeneity, with significant reductions in radiologist workload. Nevertheless, limitations exist, such as lack of model generalizability and interpretability. Conclusions: AI is revolutionizing oncologic radiology by improving diagnostic precision, reducing radiologist workload and contributing to personalized medicine. Future advancements should focus on refining model generalizability, integrating AI into clinical workflows and addressing ethical, legal and interpretability challenges to maximize patient benefit.