Advancing multi-scale network and agent-based computational lung models: potential for personalized prediction of disease progression
OA Version
Citation
Abstract
The lung is a structurally and mechanically complex organ that is susceptible to many diseases that cause permanent lung damage, such as emphysema and pulmonary fibrosis. There has been an advent of computational models of lung diseases to better understand their underlying mechanisms. To replicate the mechanical behaviors seen in the lungs, spring network models have been utilized on both the alveolar scale and the full-lung scale; additionally, agent-based models have been utilized to apply biological behaviors to spring network models. The union of these two model types — the structural and the biological — has led to important breakthroughs in understanding pathogenesis of lung diseases. However, these models have been focused on uniform tessellating shapes which do not capture the complex structural heterogeneity seen in the lungs. Additionally, these networks tend to be generic as well as esoteric, limiting broader application. There is therefore a need for a more robust means of creating subject-specific models and interpreting the results. The main thesis of this work is therefore that subject-specific spring network models can be utilized to understand the underlying mechanisms of lung pathologies, which in turn can be used to predict disease progression. This work outlines the development of non-uniform and ultimately subject-specific models of lung tissue using spring network model and agent-based model hybrids. These models are then compared to and validated by physiological data, and show complex emergent behavior consistent with real lung diseases. In conclusion, these non-uniform and subject-specific models are far more robust at capturing and understanding existing lung structural pathologies and show great promise to predict disease progression on a personalized basis.
Description
2025