Background
Lung cancer screening increasingly relies on risk-based management of pulmonary nodules. Recent artificial intelligence (AI) models have demonstrated substantially improved performance in distinguishing malignant from benign nodules, particularly for indeterminate nodules that are frequently encountered in screening practice. Despite these advances, current screening guidelines still rely on conventional risk models and size-based management strategies.
Problem description
Although AI can accurately estimate the probability that a pulmonary nodule is malignant, these risk scores are difficult to use directly in clinical practice. Screening programs require clear management recommendations, such as routine annual screening, short-term follow-up, or immediate diagnostic work-up. At present, clinically validated thresholds that translate AI-derived malignancy probabilities into actionable management decisions are lacking.
Project summary
This project aims to define clinically actionable AI-based malignancy risk thresholds for pulmonary nodule management in lung cancer screening. Using data from the landmark European lung cancer screening trials NELSON, DLCST, and MILD, we will evaluate how AI-derived malignancy probabilities can be translated into four management pathways: routine annual screening, 6‑month follow-up CT, 3‑month follow-up CT, or immediate diagnostic work-up.
The proposed AI-based strategy will be compared with current guideline-based approaches, including Lung-RADS v2022 and the latest ESTI nodule management recommendations. We will also investigate whether different thresholds are needed for specific subgroups, such as different nodule sizes, nodule types, and demographic characteristics. Finally, the study will quantify the impact on healthcare resources and costs, including the number of follow-up CT examinations and diagnostic procedures.
By translating AI predictions into practical management recommendations, the project aims to facilitate the adoption of AI in lung cancer screening, reduce unnecessary follow-up examinations, and support efficient implementation of screening programs across Europe.
Funding
This project is co-funded through the ESTI Research Support Programme and is conducted in collaboration with European lung cancer screening experts.
