Noise2NAKO(AI): "AI Methods linking Environment and Health - a large-scale cohort application"
Sophisticated spatial and spatio-temporal exposure models are urgently needed to better reflect real-life exposures and to comprehensively determine and understand the long-term impact of environmental factors on health. Furthermore, advanced statistical and data science approaches are needed to elucidate and understand the complex interplay between the environment and population health. Currently available models are hampered by the trade-off between complexity and interpretability as well as the biased nature of population-based cohort data. This project aims on solving these challenges by developing data science methods in the domains of Artificial Intelligence (AI) and Machine Learning (ML) to advance currently available noise maps, improve the quantification of noise impacts on health and delineate the complex interplay between environmental, contextual and individual socio-economic and health data. Methodically, three objectives are
- the extension of 2017 traffic noise maps (currently only available for larger agglomerations and in the vicinity of major roads, railways and airports) by the application of ML methods instead of physical modelling approaches.
- Development and systematic testing of deep learning approaches to link noise maps with neighborhood information of the German national cohort to identify vulnerable clusters in terms of noise and neighborhood for the risk of hypertension and prediction of these clusters across entire Germany.
- Extension of this prediction model with individual information from NAKO participants to investigate the additional influence of individual risk factors for hypertension exploring and promoting AI/ML and interpretable approaches.
