Artificial intelligence in thoracic imaging—a new paradigm for diagnosing pulmonary diseases: a narrative review
- Author:
Gong Yong JIN
1
Author Information
- Publication Type:Focused Issue of This Month
- From:Journal of the Korean Medical Association 2025;68(5):288-300
- CountryRepublic of Korea
- Language:Korean
- Abstract: This review explores the current applications and future prospects of artificial intelligence (AI) in thoracic imaging, with a particular focus on chest radiography (chest X-ray, CXR) and computed tomography (CT).Current Concepts: Recently developed CXR AI algorithms have improved the efficiency, accuracy, and consistency of radiologists' routine clinical workflows by assisting in the detection of a wide range of thoracic diseases on CXR. These AI systems demonstrate diagnostic performance comparable to that of radiology residents who have limited interpretive experience. Furthermore, generative CXR AI technologies are capable of not only automatically detecting abnormalities such as pulmonary nodules, pneumonia, pneumothorax, and tuberculosis, but also generating radiology reports. These advancements represent a paradigm-shifting innovation that may significantly alter the current landscape of CXR interpretation in thoracic radiology. Although performance varies depending on the specific algorithm and dataset, AI applied to low-dose chest CT has demonstrated diagnostic accuracy ranging from 0.81 to 0.98 for nodule detection and malignancy assessment, with sensitivity ranging from 0.88 to 0.99 and specificity from 0.82 to 0.93. Incorporating AI as a second reader in CT interpretation can reduce reading time by approximately 20%, while also improving sensitivity for pulmonary nodule detection by 5% to 20% and malignant nodule diagnosis by 3% to 15%.Discussion and Conclusion: Both CXR AI and chest CT AI streamline image interpretation by assisting with simple and repetitive tasks. Simultaneously, they provide novel diagnostic insights that are expected to influence and potentially reshape the interpretative patterns of radiologists in the near future.
