AI, Imaging, and Smarter Screening Drive Early Lung Disease Detection

Artificial intelligence (AI)–supported spirometry readings, acoustic biomarkers captured through a smartphone microphone, CT-based scoring of mucus plugs invisible to the naked eye, and national lung cancer screening programs that are starting to shift diagnoses to earlier, curable stages were all presented as evidence that early detection and prevention are moving from research concept to clinical practice, according to sessions on new tools for lung disease detection at the European Respiratory Society (ERS) Congress.

Digital Tools for Secondary Prevention

Amy Hai Yan Chan, PharmD, head of the School of Pharmacy and associate professor at the University of Auckland, opened the session by framing the problem digital tools are meant to solve: patients often go undiagnosed for years and are seen in clinic only every few months, leaving clinicians with little visibility into what happens in between. Chan pointed to an Australian study of AI-assisted interpretation of spirometry in more than 1000 primary care patients that found 84% sensitivity and 86% specificity compared with clinician diagnosis,¹ as well as a randomized controlled trial of 234 clinicians showing that AI-supported decision software improved diagnostic accuracy but not clinician confidence.² She also cited the 2026 GOLD guidelines for chronic obstructive pulmonary disease (COPD), which now flag emerging alternatives to spirometry, including respiratory sound analysis, oscillometry, and volatile organic compound detection, as under active investigation.³

Chan reviewed several machine learning risk models built on primary care records, including one that predicted 1-year asthma attack risk (with high-risk patients experiencing exacerbations at a 1-in-3 rate, versus 1-in-16 among those flagged low-risk),⁴ a model aimed at detecting COPD earlier in patients aged 20 to 50 years,⁵ and a "target COPD score" using age, smoking history, dyspnea, and prescription patterns to flag likely undiagnosed COPD.⁶ She also referenced the INCA-SUN (NCT02307669) study, in which biofeedback from digital inhalers improved adherence more than traditional education,⁷ alongside a US study using inhaler GPS data to build neighborhood-level pollution and exacerbation risk maps.⁸

Technology alone isn't enough, Chan cautioned. "No technology works in isolation," she said. "They still need clinicians, and they still need humans... It can't be a top-down approach, and we need to make sure it fits in the care pathway."

She pointed to cases where that redesign didn’t happen, such as an AI tool that sped up radiologist reporting in a lung cancer imaging trial but did not change time to diagnosis or referral rates⁹; a telemonitoring study in which patients had more consultations overall, more outpatient visits, and more exacerbations flagged than those receiving conventional care¹⁰; and a systematic review of more than 100 respiratory apps that found only 3% meaningfully empowered patients to self-monitor.¹¹

Imaging Biomarkers Move From Diagnostic to Theranostic

Arnaud Bourdin, MD, PhD, head of pulmonology and a professor of respiratory medicine at the University of Montpellier's Arnaud de Villeneuve Hospital in France, traced the evolution of chest CT from a purely diagnostic tool to one that can quantify disease and predict treatment response. Using AI-based scoring, Bourdin showed CT scans in which more than 20 airways were occluded by mucus plugs that would otherwise have looked unremarkable to the eye and described how mucus plug volume declines in patients responding to biologic therapy, including in cystic fibrosis patients on CFTR triple therapy.

He cited the VESTIGE study (NCT04400318), in which patients with high baseline mucus plug scores saw substantially greater benefit from dupilumab than those with minimal mucus plugging, suggesting the score could have theranostic value,¹² as well as a study of an anti–IL-33 antibody that reduced mucus plug scores in COPD by week 28.¹³

Bourdin argued that access to imaging isn't the bottleneck to wider use of these biomarkers—standardizing how scans are scored is, given the scale of CT scanning already performed worldwide each year.

"We are missing this kind of information," he said. "If we were able to score and provide some clear values of that recording in a folder, we'll probably be able to consider that this is very strong, and this may increase equity around the world, especially with AI tools."

Lung Cancer Screening Moves From Pilots to Programs

Torsten Gerriet Blum, MD, MBA, senior physician at HELIOS Klinikum Emil von Behring and a professor at Medical School Berlin, discussed the evidence base for lung cancer screening, noting that 10 randomized controlled trials, including the 2 largest—NELSON (ISRCTN63545820) and National Lung Screening Trial (NCT00047385)—showed a reduction in lung cancer mortality tied to earlier-stage diagnosis and higher curative resection rates. Blum said implementation is now visible in registry data from the United States and in Taiwan's population-based screening program, where incidence curves for early- and late-stage disease have crossed.^14,15

In Europe, he highlighted the EU4Health-funded SOLACE project, which has produced a comprehensive guideline framework expected to feed into a forthcoming European Union (EU) lung cancer screening guideline, and noted that national programs are now running or launching in Croatia, the Czech Republic, England, and, as of October, Poland.

“Lung cancer screening is ready for prime time,” Blum said. “There is no doubt about that, but we can and do optimize the eligibility inclusion criteria [and] we will go to personal risk-adapted screening, defining better intervals in the screening.”

Translating Detection Into Everyday Care

Joanna Chorostowska-Wynimko, MD, PhD, president of ERS, and head of the Department of Genetics and Clinical Immunology at the National Institute of Tuberculosis and Lung Diseases in Poland, drew on SOLACE's 13 pilot studies across 15 EU countries to outline what makes screening programs work in practice, not just on paper. She cited UK data showing that of the invitees who responded to a lung cancer screening invitation, the highest response (34%) came from less deprived, lower-risk populations, while those from more deprived, higher-risk backgrounds were the least likely to respond.¹⁶ Older age, current smoking, female sex, and lower socioeconomic status were all associated with lower uptake.

She described community-based strategies—including a mobile CT pilot in rural Poland that screened more than 3000 people and a Roma community ambassador program in Hungary¹⁷—as more effective than hospital-based recruitment alone.

Chorostowska-Wynimko also pointed to Czechia's model, in which general practitioners refer patients directly to respiratory clinics for screening and follow-up.

“We should appreciate that this is much more than just nodules,” she said. “This is chest screening,” which is designed to capture COPD and lung fibrosis alongside lung cancer, rather than a single disease.

References

1. Sunjaya A, Edwards GD, Harvey J, et al. Validation of artificial intelligence spirometry diagnostic support software in primary care: a blinded diagnostic accuracy study. ERJ Open Res. 2025;11(5):00116-2025. doi:10.1183/23120541.00116-2025

2. Doe G, Banya W, Edwards GD, et al. AI-assisted spirometry interpretation in primary care: a randomized controlled trial. NEJM AI. 2025;2(8). doi:10.1056/AIoa2400804

3. Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: 2026 report. Published 2026. Accessed September 17, 2026. https://goldcopd.org/wp-content/uploads/2026/01/GOLD-REPORT-2026-v1.3-8Dec2025_WMV2.pdf

4. Tibble H, Sheikh A, Tsanas A. Development and validation of a machine learning risk prediction model for asthma attacks in adults in primary care. NPJ Prim Care Respir Med. 2025;35(1):24. doi:10.1038/s41533-025-00428-8

5. Liu G, Hu J, Yang J, Song J. Predicting early-onset COPD risk in adults aged 20–50 using electronic health records and machine learning. PeerJ. 2024;12:e16950. doi:10.7717/peerj.16950

6. Haroon S, Adab P, Riley RD, Fitzmaurice D, Jordan RE. Predicting risk of undiagnosed COPD: development and validation of the TargetCOPD score. Eur Respir J. 2017;49(6):1602191. doi:10.1183/13993003.02191-2016

7. Mac Hale E, Greene G, Mulvey C, et al; INCA research team. Use of digital measurement of medication adherence and lung function to guide the management of uncontrolled asthma (INCA Sun): a multicentre, single-blinded, randomised clinical trial. Lancet Respir Med. 2023;11(7):591-601. doi:10.1016/S2213-2600(22)00534-3

8. Barrett M, Combs V, Su JG, Henderson K, Tuffli M; AIR Louisville Collaborative. AIR Louisville: addressing asthma with technology, crowdsourcing, cross-sector collaboration, and policy. Health Aff (Millwood). 2018;37(4):525-534. doi:10.1377/hlthaff.2017.1315

9. Woznitza N, Smith L, Rawlinson J, et al. AI-based chest X-ray prioritization in the lung cancer diagnostic pathway: the LungIMPACT randomized controlled trial. Nat Med. 2026;32(5):1737-1744. doi:10.1038/s41591-026-04253-5

10. Roubos LAC, Westland H, Hulstein-Brink NL, Visser RC, van den Berg JWK, Leenen JPL. Real-world comparison of telemonitoring versus conventional care in patients with chronic obstructive pulmonary disease and those with asthma—impact on clinical outcomes and patient characteristics: retrospective cohort study. J Med Internet Res. 2025;27:e66743. doi:10.2196/66743

11. Huckvale K, Car M, Morrison C, Car J. Apps for asthma self-management: a systematic assessment of content and tools. BMC Med. 2012;10:144. doi:10.1186/1741-7015-10-144

12. Castro M, Papi A, Porsbjerg, et al. Effect of dupilumab on exhaled nitric oxide, mucus plugs, and functional respiratory imaging in patients with type 2 asthma (VESTIGE): a randomised, double-blind, placebo-controlled, phase 4 trial. Lancet Respir Med. 2025;13(3):208-220. doi:10.1016/S2213-2600(24)00362-X

13. Nordenmark L, Guller P, Reid F, et al. S91 Tozorakimab (anti-IL-33 mAb) reduces mucus plugging in COPD: an imaging sub-study in the FRONTIER-4 phase 2a COPD trial. Thorax. 2024;79:A67-A68. doi:10.1136/thorax-2024-BTSabstracts.97

14. Vachani A, Carroll N, Simoff M, et al. Stage migration and lung cancer incidence after initiation of low-dose CT screening. J Thorac Oncol. 2022;17(12):1355-1364. doi:10.1016/j.jtho.2022.08.011

15. Association of computed tomographic screening promotion with lung cancer overdiagnosis among Asian women. JAMA Intern Med. 2022;182(3):283-290. doi:10.1001/jamainternmed.2021.7769

16. Ali N, Lifford KJ, Carter B, et al. Barriers to uptake among high-risk individuals declining participation in lung cancer screening: a mixed methods analysis of the UK Lung Cancer Screening (UKLS) trial. BMJ Open. 2015;5(7):e008254. doi:10.1136/bmjopen-2015-008254

17. European Lung Foundation. SOLACE: Strengthening the Screening of Lung Cancer in Europe. Accessed September 17, 2026. https://europeanlung.org/solace/