Air Quality and Breast Malignancy Risk Stratification in the Mediterranean: Integrating Copernicus Reanalysis Data with Mammographic Feature

Air pollution is a major environmental determinant of human health, and evidence suggests that chronic exposure to atmospheric pollutants may influence breast cancer risk. This is relevant in the Mediterranean region, where urban emissions, industrial sources, regional transport, Saharan dust intrusions and climate-related stressors create a complex exposure scenario. However, air quality indicators are rarely incorporated into malignancy prediction models. This study assessed whether long-term exposure estimates derived from Copernicus Atmosphere Monitoring Service (CAMS) reanalysis data could provide complementary information for breast lesion malignancy stratification in a Mediterranean screening population. We retrospectively analysed mammographic and clinical data from 906 women undergoing breast cancer screening. Lesions were classified as benign (BI-RADS B2) or malignant (BI-RADS B5). Residential zip codes were linked to CAMS gridded concentration fields to estimate individual exposure to nitrogen dioxide (NO₂), fine particulate matter (PM₂.₅), coarse particulate matter (PM₁₀) and ozone (O₃). For each pollutant, annual mean concentrations and cumulative exposure over the three years preceding diagnosis were calculated. These environmental metrics were integrated with demographic information and mammographic descriptors, including lesion morphology, margins and breast density patterns. To reduce model complexity and limit overfitting, variables were screened using univariate ANOVA F-tests, retaining predictors with p < 0.05. Selected features were then used to train a feed-forward neural network for classification. Performance was evaluated on validation data and compared with models excluding environmental exposure variables. Correlations among pollutants were examined to assess collinearity and confounding. The integrated model achieved a ROC-AUC of 0.78, with balanced accuracy and weighted F1-score equal to 0.73. Radiological features remained the strongest predictors of malignancy, particularly spiculated margins and irregular lesion shape. Nevertheless, cumulative NO₂ and PM₂.₅ exposure retained independent statistical significance and contributed to model discrimination. Removing highly correlated air quality variables reduced apparent predictive gain but improved model stability and interpretability, highlighting the need for cautious exposure selection in observational environmental health studies. These findings suggest that long-term air pollution exposure, quantified through CAMS atmospheric reanalysis products, may provide a modest but consistent contribution to breast lesion malignancy prediction when combined with mammographic features. Although no causal inference can be drawn, the study supports the feasibility of integrating atmospheric composition data, medical imaging and machine learning within a transdisciplinary environmental health framework. Larger cohorts, finer geocoding and external validation are needed to confirm these associations.

References

[1] White AJ, Bradshaw PT, Hamra GB. Air pollution and breast cancer: a review. Curr Epidemiol Rep. 2018;5(2):92-100. doi:10.1007/s40471-018-0143-2.

[2] Praud D, Deygas F, Amadou A, Bouilly M, Turati F, Bravi F, et al. Traffic-related air pollution and breast cancer risk: a systematic review and meta-analysis. Cancers. 2023;15(3):927. doi:10.3390/cancers15030927.

[3] Inness A, Ades M, Agustí-Panareda A, Barré J, Benedictow A, Blechschmidt AM, et al. The CAMS reanalysis of atmospheric composition. Atmos Chem Phys. 2019;19:3515-3556. doi:10.5194/acp-19-3515-2019.

[4] Fiore M, Palella M, Ferroni E, Miligi L, Portaluri M, Marchese CA, et al. Air pollution and breast cancer risk: an umbrella review. Environments. 2025;12:289. doi:10.3390/environments12050153.