Back to all papers

Integration of closed-loop fully-automated AI model with clinician assessment for lung nodule stratification: A multi-reader study.

August 27, 2026pubmed logopapers

Authors

Taha A,Kheir F

Affiliations (2)

  • Division of Pulmonary and Critical Care Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, 02114, United States. Electronic address: [email protected].
  • Division of Pulmonary and Critical Care Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, 02114, United States.

Abstract

Computed tomography (CT)-based artificial intelligence (AI) risk models improve malignancy prediction, but their incremental value when integrated into clinician workflows has not been studied. We aimed to assess the sequential application of Bronchosolve (AI) in nodule risk stratification. We performed a fully crossed, multi-reader, multi-case retrospective study of 296 chest CT scans from screening and incidentally detected lung nodules under three conditions: clinician, Bronchosolve (AI), and AI-aided clinician interpretation. Primary outcomes were area under the receiver operating characteristic curve (AUC) and accuracy. Secondary outcomes included sensitivity, specificity, and net reclassification improvement (NRI), with prespecified analysis of intermediate-risk nodules. Mean AUC increased from 0.84 (95% CI, 0.83-0.86) for clinicians to 0.87 (95% CI, 0.86-0.88) for Bronchosolve and 0.88 (95% CI, 0.86-0.89) for AI-aided clinician interpretation. Overall, readers' accuracy improved from 74.7% to 77.7% with AI assistance (p < 0.01). In the intermediate-risk subgroup, the sequential AI-assisted workflow achieved 88.2% sensitivity and 52.3% specificity, compared with 84.8% sensitivity for Bronchosolve alone. NRI for the full cohort was 0.19, driven predominantly by correct upward reclassification of malignant nodules (event NRI = 0.10). In the full cohort, there were minimal reclassification changes among initially Low- and High-Risk cases. Bronchosolve primarily reclassified Intermediate-Risk nodules into binary Low- or High-Risk final categories. Of 969 Intermediate-Risk reader case assessments, 47.8% were reassigned to Low-Risk (71.1% benign) and 52.2% to High-Risk (59.7% malignant), demonstrating clinically meaningful risk resolution. Integrating Bronchosolve with clinician assessment improved discrimination and clinically relevant reclassification of intermediate risk lung nodules.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.