Multiscale lung ventilation modeling in health and disease

Nicolas Pozin 1
1 REO - Numerical simulation of biological flows
LJLL - Laboratoire Jacques-Louis Lions, UPMC - Université Pierre et Marie Curie - Paris 6, Inria de Paris
Abstract : The lung is an organ of the respiratory system. It supplies the organism with oxygen, which is necessary to the metabolism, and ensures carbon dioxide release. Humans have two lungs. The latter contain a tree, referred to as tracheo-bronchial, through which the air flows. It supplies a porous region, the parenchyma, where gas exchanges with the blood take place. Some pathologies affect the tree structure or the parenchyma integrity. They can induce ventilation defects or increased respiratory efforts. In vivo-studies are complex and mathematical modeling can provide some insights on the lung behavior, the pathologies’ impacts or the efficiency of treatments. I n the first part of this thesis, we propose a ventilation model of the lung based on a mechanical description. A 0D model of the tracheo-bronchial tree is strongly coupled to a continuous 3D model of the parenchyma. An efficient implementation is proposed. We numerically show the influence of chosen boundary conditions as well as tree or parenchyma alterations on the ventilation distribution. Results are compared with those provided by a model, often used in the literature, in which the parenchyma is described as a set of mechanically independent compartments. Significant differences are obtained, in particular in pathological cases. In a second part, we use the 0D tree - 3D parenchyma coupled model to investigate how breathing gas mixtures less dense than air would reduce efforts and ensure a better ventilation. To that end, we build an asthmatic tree model based on a literature review. We propose lines of thoughts on why, as observed in clinical trials, helium-oxygen mixtures have varying efficiency. The broncho-constriction distribution, i.e. the position and severity of constrictions, is a major explaining factor. In the next part, we develop an approach to get insights on severe constrictions distribution based on the analysis of dynamic lung ventilation images. To do so, the 0D tree - 3D parenchyma coupled model is used along with a machine learning technique. Finally, two prospective works are presented. First, we propose extensions to the ventilation models introduced in the first part as a step towards spirometry modeling, a standard test to assess respiratory pathologies. The last study is part of a global perspective that aims at getting insights on the lung geometry based on simple measurements on the patient’s body. The final objective is to determine, in a simple way, with little effort, input data for the ventilation models.
Complete list of metadatas

Cited literature [146 references]  Display  Hide  Download
Contributor : Nicolas Pozin <>
Submitted on : Monday, January 15, 2018 - 11:37:21 PM
Last modification on : Wednesday, March 27, 2019 - 1:18:54 AM
Long-term archiving on : Monday, May 7, 2018 - 9:00:00 PM


Files produced by the author(s)


  • HAL Id : tel-01684990, version 1


Nicolas Pozin. Multiscale lung ventilation modeling in health and disease. Modeling and Simulation. Paris 6, 2017. English. ⟨tel-01684990v1⟩



Record views


Files downloads