The LUCIA project: understanding determinants to improve lung cancer screening and early detection
Abstract
Background: Lung cancer (LC) is the leading cause of cancer-related deaths globally. The LUCIA project (Agreement n° 101096473) is dedicated to advance LC screening and prevention by discovering and understanding risk factors contributing to LC development. We use digital tools and AI to integrate multimodal data from wide-population and precision screening cohorts. Linking disparate data layers, we aim to develop a validated risk engine for more accurate, individualized risk stratification in LC screening programs.
Methods: LUCIA is a multinational study integrating retrospective Electronic Health Record (EHR) cohorts with a prospective clinical validation study. We included >50,000 individuals from longitudinal EHR datasets for the retrospective study and 3,029 participants in the prospective study enrolled in 3 phases. We use data harmonization to the OMOP Common Data Model, advanced AI modeling using Time-Aware Transformers (STraTS) for clinical time-series and hybrid models for environmental/geospatial data fusion. As intervention, we apply non-invasive sensing devices. Our primary outcomes include model predictive performance (AUROC/AUPRC) for LC risk, diagnostic sensitivity/specificity of sensing technologies, and identification of implementation.
Results: Preliminary results demonstrate a high predictive and diagnostic performance across the LUCIA toolbox. We have validated a multimodal risk engine integrating clinical and environmental data to achieve high predictive accuracy (AUROC 0.75), outperforming traditional screening models. By demonstrating that non-invasive breath and wearable sensors provide effective triage with 80–87% sensitivity, we support a low-cost primary care screening layer to reduce unnecessary imaging. These technical achievements are backed by the harmonization of 90% of consortium data to OMOP standards.
Conclusions: LUCIA has transitioned a holistic AI toolbox to the public health domain to enable proactive, precision LC prevention.
Key messages: • The LUCIA project (2023-2026) works to validate risk factors beyond smoking history in lung cancer development and to enable proactive, precision lung cancer prevention.
• The LUCIA project results demonstrate the effectiveness of non-invasive breath and wearable sensors and support a low-cost primary care screening layer to reduce unnecessary imaging.
Related
- Potential to improve early detection of lung cancer through expansion of focused and opportunistic screening approaches: a scoping review focusing on potential expansion of screening risk factors and adoption of artificial intelligence models.
- Failed quality improvement project to improve adherence to CAM-ICU and RASS documentation in a tertiary intensive care unit: a cautionary tale.
- Understanding Cardiac Anatomy and Imaging to Improve Safety of Procedures
- From digital reminders to artificial intelligence–enabled understanding: can personalized treatment predictions improve therapeutic adherence?
- Improving obstructive sleep apnea screening in patients with atrial fibrillation (SLEEP-AF project)
BibTeX
@article{nabbe2026lucia,
title = {{The LUCIA project: understanding determinants to improve lung cancer screening and early detection}},
author = {M Nabbe and A Garin-Muga and E Arana-Arri and B Polis},
journal = {The European Journal of Public Health},
year = {2026},
month = oct,
doi = {10.1093/eurpub/ckag169.608},
url = {https://doi.org/10.1093/eurpub/ckag169.608},
note = {CC BY-NC 4.0},
}
RIS
TY - JOUR
TI - The LUCIA project: understanding determinants to improve lung cancer screening and early detection
AU - M Nabbe
AU - A Garin-Muga
AU - E Arana-Arri
AU - B Polis
PY - 2026
DA - 2026/10/01
JO - The European Journal of Public Health
DO - 10.1093/eurpub/ckag169.608
UR - https://doi.org/10.1093/eurpub/ckag169.608
AB - Background: Lung cancer (LC) is the leading cause of cancer-related deaths globally. The LUCIA project (Agreement n° 101096473) is dedicated to advance LC screening and prevention by discovering and understanding risk factors contributing to LC development. We use digital tools and AI to integrate multimodal data from wide-population and precision screening cohorts. Linking disparate data layers, we aim to develop a validated risk engine for more accurate, individualized risk stratification in LC screening programs. Methods: LUCIA is a multinational study integrating retrospective Electronic Health Record (EHR) cohorts with a prospective clinical validation study. We included >50,000 individuals from longitudinal EHR datasets for the retrospective study and 3,029 participants in the prospective study enrolled in 3 phases. We use data harmonization to the OMOP Common Data Model, advanced AI modeling using Time-Aware Transformers (STraTS) for clinical time-series and hybrid models for environmental/geospatial data fusion. As intervention, we apply non-invasive sensing devices. Our primary outcomes include model predictive performance (AUROC/AUPRC) for LC risk, diagnostic sensitivity/specificity of sensing technologies, and identification of implementation. Results: Preliminary results demonstrate a high predictive and diagnostic performance across the LUCIA toolbox. We have validated a multimodal risk engine integrating clinical and environmental data to achieve high predictive accuracy (AUROC 0.75), outperforming traditional screening models. By demonstrating that non-invasive breath and wearable sensors provide effective triage with 80–87% sensitivity, we support a low-cost primary care screening layer to reduce unnecessary imaging. These technical achievements are backed by the harmonization of 90% of consortium data to OMOP standards. Conclusions: LUCIA has transitioned a holistic AI toolbox to the public health domain to enable proactive, precision LC prevention. Key messages: • The LUCIA project (2023-2026) works to validate risk factors beyond smoking history in lung cancer development and to enable proactive, precision lung cancer prevention. • The LUCIA project results demonstrate the effectiveness of non-invasive breath and wearable sensors and support a low-cost primary care screening layer to reduce unnecessary imaging.
SN - 1101-1262
SN - 1464-360X
ER -