IMPACT OF ARTIFICIAL INTELLIGENCE ALGORITHMS ON THE DETECTION AND CHARACTERIZATION OF PULMONARY NODULES IN CHEST COMPUTED TOMOGRAPHY

Introduction: Pulmonary nodules detected on chest computed tomography represent a common and clinically relevant finding, with significant implications for early lung cancer diagnosis and patient management. The increasing volume of imaging studies has intensified interest in artificial intelligence algorithms as tools to improve detection accuracy, reduce interobserver variability, and optimize workflow efficiency in thoracic radiology. Objective: The main objective of this systematic review is to evaluate the impact of artificial intelligence algorithms on the detection and characterization of pulmonary nodules in chest computed tomography. Secondary objectives include assessing diagnostic accuracy compared with human readers, analyzing performance across different nodule sizes and types, evaluating effects on radiologist workflow, examining generalizability across populations and scanners, and summarizing current limitations and regulatory considerations. Methods: A systematic search was conducted across PubMed, Scopus, Web of Science, Cochrane Library, LILACS, ClinicalTrials.gov, and the International Clinical Trials Registry Platform. Studies published within the last five years were prioritized, with extension up to ten years if necessary, and eligibility criteria focused on human studies evaluating artificial intelligence–based algorithms for pulmonary nodule detection or characterization on chest computed tomography. Data synthesis followed PRISMA recommendations with structured qualitative analysis. Results and Discussion: A total of 20 studies met the inclusion criteria and were included in the final review. Overall, artificial intelligence algorithms demonstrated high sensitivity for pulmonary nodule detection, particularly for small and subsolid nodules, and showed consistent improvements in reader performance when used as decision-support tools. Variability in study design, reference standards, and outcome reporting contributed to moderate heterogeneity, but most studies supported a complementary role for artificial intelligence in clinical practice rather than full replacement of expert interpretation.Conclusion: Current evidence suggests that artificial intelligence algorithms can enhance the detection and characterization of pulmonary nodules on chest computed tomography, especially when integrated into radiologist workflows. While promising, their clinical implementation requires careful validation, transparency, and alignment with evidence-based guidelines to ensure safe and effective use.