AI Enhances Lung Nodule Detection Without Reducing Interpretation Time

Overview

In a randomized clinical trial involving 911 asymptomatic individuals undergoing low-dose chest CT, AI-assisted interpretation significantly increased lung nodule detection rates but did not reduce the time required for scan interpretation. Although AI led to more follow-up imaging recommendations, no lung cancer diagnoses were made during a median follow-up of seven months.

Background

Lung nodule detection on low-dose chest computed tomography (LDCT) is critical for early identification of potential lung cancer. Artificial intelligence (AI) tools have been developed to assist radiologists by improving detection accuracy and efficiency. However, the impact of AI on interpretation time and clinical outcomes remains uncertain. This study evaluated the use of an AI-based lung nodule evaluation tool integrated into the picture archiving and communication system during routine health checkups in South Korea.

Data Highlights

Key Findings

• AI-assisted interpretation increased detection of Lung-RADS–positive nodules from 10% to 17% of scans.
• Overall nodule detection improved from 33% without AI to 53% with AI assistance.
• The number of nodules detected per scan was higher with AI, especially for nodules measuring 4 to 8 mm.
• Interpretation time per examination was not significantly different between AI and standard groups (187 vs 172 seconds).
• Follow-up imaging recommendations were more frequent with AI (15% vs 7%).
• No lung cancer diagnoses occurred in either group during a median follow-up of approximately seven months.

Clinical Implications

AI integration into LDCT interpretation can enhance detection of clinically actionable lung nodules without increasing interpretation time, potentially improving early identification of suspicious findings. However, increased detection may lead to more follow-up imaging, which requires careful consideration to avoid unnecessary procedures. Clinicians should weigh the benefits of improved sensitivity against the lack of demonstrated short-term clinical impact and potential workflow changes.

Conclusion

Use of AI in LDCT interpretation significantly improves lung nodule detection rates without reducing interpretation time, though the short-term clinical benefit remains uncertain. Further studies with longer follow-up and diverse clinical settings are needed to clarify the impact on patient outcomes.

References

1. Hwang EJ et al. 2024 -- AI May Improve Lung Nodule Detection in Low-Dose Chest CT