The case for early detection. Lung cancer is the leading cause of cancer death worldwide. When diagnosed based on symptoms, most cases are already at an advanced, inoperable stage. Lung cancer screening with low-dose CT (LDCT) detects tumors before symptoms appear, enabling earlier treatment and significantly better survival outcomes.
Evidence from major clinical trials. Two landmark randomized controlled trials - NLST in the United States and NELSON in Europe - demonstrated that LDCT screening of high-risk populations reduces lung cancer-specific mortality. Additional European trials (MILD and LUSI) confirmed these findings, establishing a strong evidence base for population-level screening programs.
Who benefits from screening. Current European screening programs focus on high-risk individuals: current or former smokers aged 50-74 with at least 20-30 pack-years of smoking history who quit within the last 10-15 years. Refining these eligibility criteria to capture additional high-risk groups remains an active area of research.
Growing momentum across Europe. Following a 2022 European Council recommendation, several countries including Poland, Croatia, and the United Kingdom have launched or are setting up national lung cancer screening programs. The EU4Health-funded SOLACE project aims to strengthen and harmonize screening implementation across European member states.
The challenge of nodule management. Low-dose CT scans frequently detect small lung nodules - the vast majority of which are harmless - but some are early-stage cancers. Deciding which nodules need immediate investigation, which need monitoring, and which can be safely ignored is a complex clinical challenge that requires standardized guidelines.
Negative results: very low cancer risk. A negative screen result means the estimated one-year risk of a detected nodule being lung cancer is very low. This category represents 80-85% of all screening participants, and they return to annual screening without any additional follow-up scans.
Indeterminate results: low to moderate risk. An indeterminate result requires a shorter follow-up CT scan - at 1, 3, or 6 months depending on the guideline - to determine if the nodule is growing, shrinking, or stable. Indeterminate nodules are found in roughly 10-20% of screening participants and require active monitoring without immediate invasive workup.
Positive results: high cancer risk. A positive screen result indicates a high estimated probability of cancer. These cases require further diagnostic workup and referral to a multidisciplinary team meeting to plan the most appropriate management, which may include tissue biopsy, PET scan, or surgical resection.
Morphological classification. The most widely used approach - exemplified by the US Lung-RADS v2022 guideline - classifies nodules based on detailed visual features observed by a radiologist. Key characteristics include nodule density (solid versus ground-glass), size, shape, and the presence of suspicious features like spiculation or irregular borders.
Model-based risk calculation. Some guidelines, including the International Lung Screen Trial and the British Targeted Lung Health Check, use mathematical models such as the Brock model. The radiologist enters individual features (age, family history, nodule location, emphysema presence) into the model, which automatically computes a cancer probability. This approach has a higher negative predictive value because it incorporates non-imaging clinical factors.
Deep learning-based classification. The newest approach feeds CT images directly into deep convolutional neural network models that automatically compute a malignancy probability. While no current clinical guidelines have fully integrated these models, multiple commercial and academic solutions exist and are being prospectively tested. Deep learning models have already been shown to outperform traditional multivariable risk models in some studies.
Advantages and trade-offs. Morphological classification works well for most typical nodules, while model-based and AI approaches reduce reader variability for complex cases. Each method has practical limitations: morphological classification requires expert pattern recognition, model-based tools require accurate data entry, and deep learning tools require validation in diverse real-world populations before widespread adoption.
Suspicious features on CT. Several CT characteristics raise concern for malignancy. Spiculation - irregular spike-like projections at the nodule margin - and signs of architectural distortion like pleural indentation or fissure displacement strongly suggest a solid malignant nodule. Bubble-like lucencies, concave margins, and narrowed adjacent vessels are additional warning signs.
Solid component size matters. For part-solid nodules (which have both solid and ground-glass components), the size of the solid component is the best predictor of invasiveness. A solid component exceeding 5 mm corresponds to minimally invasive adenocarcinoma. When the solid component exceeds 80% of the total nodule diameter, the nodule should be treated as a solid nodule for management purposes.
Atypical cysts as a new category. The 2022 Lung-RADS update introduced the 'atypical pulmonary cyst' category, recognizing that lung cysts with irregular or thickening walls that develop over time can be highly suspicious for malignancy. This category was not formally recognized in earlier guidelines.
Features that suggest benignity. Conversely, certain features indicate a nodule is almost certainly benign. Calcification patterns such as central, diffuse, or popcorn-like calcification suggest granulomas or hamartomas. Intranodular fat and smooth, well-defined borders also indicate benign lesions. Intrapulmonary lymph nodes - which make up a large proportion of small solid nodules seen at screening - can be recognized by their characteristic oval or lentiform shape, smooth margins, and location close to the pleura below the level of the carina.
The geometry problem with diameter measurements. A lung tumor must double in volume before its diameter increases by just 26%. This means that diameter-based measurements are insensitive to early growth. Additionally, manual diameter measurements carry substantial variability between different readers (up to 1.7 mm) and even within the same reader, leading to false impressions of growth or stability.
Volumetric software captures three-dimensional growth. Computer-assisted volumetry measures the full three-dimensional volume of a nodule, detecting early growth - including asymmetric growth along the z-axis - that two-dimensional measurements would miss. Volume doubling time (VDT), calculated from volumetric measurements, is the key parameter for assessing how quickly a nodule is growing and correlates with cancer aggressiveness.
Limitations of volumetric software. Despite its advantages, volumetric measurement is imperfect. Software struggles with non-solid (ground-glass) components because they have similar density to surrounding lung tissue. Nodules touching the pleura or blood vessels may have their volume overestimated because the software includes adjacent structures. Inspiration level, patient repositioning, and CT reconstruction parameters all introduce variability.
Consistency is essential. The same volumetry software should be used throughout a patient's screening follow-up, since different software products give different results. If software is changed, earlier measurements must be repeated if the change might affect clinical management. For subsolid nodules where volumetry is unreliable, a visually confirmed diameter increase of more than 1.5 mm per year is used as the threshold for growth.
Aggressiveness-guided management. The European Society of Thoracic Imaging (ESTI) developed its nodule management recommendation specifically to focus on lesion aggressiveness rather than just size thresholds. At baseline, aggressiveness is primarily determined by nodule type (solid versus subsolid) and morphological features, with suspicious features upgrading and benign features downgrading the risk category.
Solid nodule thresholds. For solid nodules at baseline, those smaller than 100 mm3 (about 6 mm effective diameter) without suspicious features are classified as negative, returning to annual screening. Nodules between 100 and 500 mm3 are indeterminate and require 3 or 6-month follow-up. Nodules 500 mm3 or larger with suspicious features are positive and require immediate workup by a multidisciplinary team.
Subsolid nodule considerations. Subsolid nodules (ground-glass or part-solid) follow different rules. A non-solid component of 30 mm or more in effective diameter is classified as indeterminate. For part-solid nodules, the size of the solid component drives management. A persistent solid component of 500 mm3 or more after follow-up requires multidisciplinary team discussion.
Growth criteria during follow-up. When assessing nodule growth, the ESTI recommends VDT thresholds that change depending on follow-up duration: less than 250 days at 3 months, less than 400 days at 6 months, and less than 500 days at 12 months indicate substantial growth and trigger a positive result. Slow-growing nodules - even when their diameter increases by more than 5 mm from baseline - are also referred to the multidisciplinary team.
AI tools already showing strong performance. More than 15 CE-certified AI algorithms for lung nodule detection and volumetric assessment are now commercially available in Europe. Studies have demonstrated that deep learning models using CT image data alone can outperform traditional multivariable risk models like the Brock model for estimating the probability that a nodule is malignant.
AI for distinguishing pre-invasive from invasive lesions. Researchers have developed deep learning and radiomics approaches to differentiate pre-invasive from invasive adenocarcinomas presenting as subsolid nodules. However, these AI tools have not yet proven superior to experienced radiologists measuring the size of the solid component - a reminder that AI must clear a high performance bar before replacing established clinical methods.
Promise of systematic AI integration. Once prospectively validated in large real-world populations, AI-based nodule management tools could reduce the number of follow-up CT scans for benign nodules and speed up identification of malignant ones. This would decrease radiation exposure, anxiety, and healthcare costs while maintaining or improving cancer detection rates.
Broader opportunities: the Big-3 diseases. Since low-dose chest CT simultaneously images the lungs, heart, and aorta, it can also detect COPD and cardiovascular disease in addition to lung cancer - the three leading causes of death in screening participants. Future LDCT programs may incorporate quantitative assessment and management guidelines for all three conditions, further increasing the value of each screening scan.
The core challenge of implementation. The evidence for LDCT lung cancer screening is now well established, but translating that evidence into effective, equitable, and efficient national programs remains challenging. Key concerns include false positives, unnecessary follow-up scans, overdiagnosis, and the risk of allowing true cancers to progress while under surveillance.
ESTI's contribution to standardization. The ESTI nodule management recommendation addresses these concerns by refining the definitions of positive, indeterminate, and negative screen results. By emphasizing lesion aggressiveness over fixed size thresholds alone, the framework aims to reduce the number of unnecessary intermediate CT scans while avoiding stage shifts in cancers that need prompt treatment.
Structured reporting for consistency. The ESTI recommendation includes a structured report template to ensure consistent communication of findings between radiologists, clinicians, and multidisciplinary teams. Standardized reporting reduces ambiguity, improves data quality, and facilitates audit and quality assurance across screening programs.
Ongoing validation and refinement. Implementation trials are underway to prospectively assess the real-world performance of the ESTI framework and to evaluate the impact of AI tools in actual clinical practice. The authors anticipate that continued evidence accumulation will allow ongoing refinement of thresholds, follow-up intervals, and the integration of AI-based decision support into routine lung cancer screening.