Comparative Performance of Artificial Intelligence and Radiologists in Detecting Lung Nodules and Breast Lesions on...

This systematic review evaluates the comparative effectiveness of artificial intelligence (AI) and radiologists in the detection of breast lesions and pulmonary nodules using mammography, computed tomography, and magnetic resonance imaging. A comprehensive literature search identified six high-quality studies, including randomized controlled trials, prospective clinical investigations, and large-scale real-world implementation studies. Although liver imaging was included in the initial search strategy, no eligible comparative studies meeting the inclusion criteria were identified; therefore, the final analysis focused exclusively on breast and lung imaging. Across these settings, AI consistently demonstrated non-inferior or superior diagnostic accuracy compared to radiologists, with additional benefits such as reduced workload, shorter assessment times, improved triage efficiency, and enhanced predictive values. While large population-based studies confirmed the scalability of AI in cancer screening, smaller diagnostic studies highlighted its ability to support radiologists in complex cases. However, heterogeneity in study design, imaging modality, and outcome reporting, as well as concerns regarding generalizability and potential bias, underscore the need for further validation. These findings support the controlled integration of AI into clinical workflows to improve diagnostic performance and healthcare efficiency.