AI-based chest X-ray prioritization in the lung cancer diagnostic pathway: the LungIMPACT randomized controlled trial.
Authors
Affiliations (15)
Affiliations (15)
- University College London Hospitals NHS Foundation Trust, London, UK.
- Lungs for Living Research Centre, University College London, London, UK.
- Leeds Cancer Research UK Clinical Trials Unit, University of Leeds, Leeds, UK.
- , Tipton, UK.
- Department of Radiology, Nottingham University Hospitals, Nottingham, UK.
- Radiology Department, University Hospitals Birmingham NHS Foundation Trust, Birmingham, UK.
- The Early Diagnosis and Detection Centre, The Royal Marsden NHS Foundation Trust, London, UK.
- Institute of Cancer Research, London, UK.
- National Heart and Lung Institute, Imperial College London, London, UK.
- Department of Thoracic Medicine, University College London Hospital, London, UK.
- Research Department of Primary Care and Population Health, University College London, London, UK.
- Radiology Department, University Hospitals of Leicester NHS Trust, Leicester, UK.
- Radiology Department, East Suffolk and North East Essex NHS Foundation Trust, Colchester, UK.
- Respiratory Medicine, Nottingham University Hospitals NHS Trust, Nottingham, UK. [email protected].
- University of Nottingham, Nottingham, UK. [email protected].
Abstract
Prioritizing artificial intelligence (AI)-detected imaging findings may reduce the time to diagnosis of lung cancer. This prospective, multicentre, randomized controlled trial tested whether immediate AI prioritization of primary care-requested chest X-rays (CXR) influenced time to computed tomography (CT) and lung cancer diagnosis, the primary outcomes. Secondary outcomes included the number of urgent suspected lung cancer referrals, incidence and stage of lung cancer, times to urgent referral and treatment, concordance between AI and radiology reports, and algorithm accuracy. AI was available in both study arms, with AI prioritization randomized by day. Of 97,731 participant CXRs, 4,405 were excluded due to data compliance issues or failure of randomization, resulting in 93,326 CXRs analyzed (45,987 and 47,339 in the prioritization 'on' or 'off' arms, respectively). A total of 13,347 CTs were identified, with 2,766 performed within 14 days of CXR. Median (interquartile range) times to CT were 53 days (17-145) and 53 days (19-141), with and without AI prioritization, corresponding to a ratio of geometric means of 0.97 (95% confidence interval (CI) = 0.93-1.02; P = 0.31). When restricted to CTs performed within 14 days of CXR, the median time to CT was 8 days (5-11) in both groups. Lung cancer was diagnosed in 558 people (0.6% of CXRs). Median times to diagnosis were 44 days (26-90) and 46 days (24-105) respectively, with a ratio of geometric means of 0.98 (95% CI = 0.83-1.16; P = 0.84). No significant differences were observed in time to lung cancer referral (14 versus 15 days; P = 0.13), time to treatment (76 versus 72.5 days; P = 0.99) or stage at diagnosis (P = 0.34). Discordance between AI and radiology reports occurred in 28,261 CXRs (30.3%) and expert radiology review identified actionable findings in 6,750 cases (23.9%). AI prioritization of CXR requested by UK primary care has no significant impact on the lung cancer pathway. Therefore, CXR AI deployments should not include worklist prioritization in this context. Future research should differentiate between primary pathway changes and the direct impact of AI. ISRCTN registration: 78987039 .