Christopher Menard
Building Better Lungs
Mentors: Dr. Mark Conradi and Dr. Jason Woods, Physics Department, Washington University inSt. Louis 

The effective diffusivity of hyperpolarized ³He gas in lungs can be measured at long diffusion distances (2-3 cm) using magnetic resonance imaging. The measured diffusivity in healthy lung is reduced by a factor of 40 from its unrestricted value of 0.9 cm²/sec (³He dilute in air). This long-range ³He diffusivity is significantly greater in the lungs of individuals with emphysema (due to the destruction of alveolar walls), making this technique a potential imaging biomarker for the severity of emphysema. The measured diffusivity of ³He in healthy volunteers, however, is still much greater than the diffusivity calculated in randomly generated computer models of the human lung, using Monte Carlo simulation. A possible reason for this discrepancy is that these simulated lungs do not correctly “fill space;” they contain regions of unrealistically densely packed acini (terminal collections of air sacs in the lung) and other regions of nearly empty space.

In order to generate a more realistic model of the lung, the simulated lungs were “evolved” using a computer program that employed an algorithm inspired by evolutionary biology. An initial population of sixteen lungs was randomly generated using the unrealistic model previously employed in the research group. The lungs with a more uniform spatial distribution of acini were used to generate new lungs by randomly switching branches from lung to lung. This “breeding” process generated eight new lungs, which were quantitatively evaluated for spatial homogeneity and then, if appropriate, replaced the lungs with a less uniform spatial distribution. Within thirty generations of this process, evolved lungs with a much more uniform spatial distribution of acini were generated.

When the diffusivity in these evolved lungs was calculated with the diffusion program, there was no statistically significant difference between the diffusivity of the un-evolved lungs and the diffusivity of the evolved lungs. This implies that the spatial inhomogeneity of the initial lungs was not responsible for the discrepancy between simulation and experimental results. Another possibility for the discrepancy between the measured value in the volunteers and the calculated value from the simulated lung structures is the existence of collateral ventilation pathways. These pathways are small holes in the tissue that connect different branches of the lung and were not implemented in this computer model. Collateral pathways would greatly decrease the distance that a molecule of ³He gas would have to diffuse to travel between the various branches of the lung and therefore increase the diffusivity. Future work should incorporate these pathways into the computer simulation.