Computational Modelling Group

Image Based Modelling of Fluid Flow through Lymph Nodes

26th September 2012
26th September 2016
Research Team
Laura Cooper
Tiina Roose, Bharathram Ganapathisubramani, Geraldine Clough

Fluid flow through lymph node. Stream tubes show fluid flow pathways, the colour indicates flow rate. The grey areas show the location of blood vessels. Scale bar = 250 µm.

The lymphatic system returns fluid to the blood stream from the tissues to maintain tissue fluid homeostasis. Lymph nodes are an important part of the immune system: they filter lymphatic fluid, are a site for transfer of immune cells to and from the circulatory system and regulate lymph protein content. Inlet and outlet flow conditions have been measured in experiments. Pathways for lymph flow have been inferred from the structure of the lymph node and by tracking florescent particles passing through the node, but due to the filtering properties of the lymph node the fluid flow may not follow the same path.

The aim of this project is to use image based modelling to investigate how the internal structure of the node affects the fluid flow pathways, which are difficult to determine experimentally. The images are used to identify the geometry and structure within the node and to determine the permeability of the lymph node interstitium to lymphatic fluid. Experimental data [1] is used to determine boundary conditions and parameters for the model. Two types of imaging data of murine lymph nodes are used for this project, selective plane illumination microscopy (SPIM) and micro computed tomography (microCT). The modelling is implemented in Comsol multiphysics, a commercial finite element analysis software.

The Iridis cluster is essential for this project. The SPIM images have the lowest spatial resolution of 5 µm. This means that to capture this level of detail for a lymph node approximately 1 mm x 1 mm x 1 mm in size requires at least 106 elements. This requires 28 GB of random access memory (RAM) to calculate the results using a direct solver, which takes only 20-30 minutes to solve on Iridis. 192 lymph node models of this size, each with six different outlet parameters, have been sent to Iridis as part of an iterative optimisation process to determine three parameters for the model. This process used over 81 hours of computational time. It would have been impossible to carry out this process on the average desktop computer, not only because of the RAM required but also as it was possible to utilises 16 processors on Iridis. This allowed the six outlet parameters to be run in parallel, so a model requiring 6 runs only took a few minutes longer than a single run. MicroCT images with a higher spatial resolution of 2.76 µm have been used of larger lymph nodes, approximately 1 mm x 1 mm x 2 mm. The meshes for these models require around 3 x 106 elements and the high memory nodes available on Iridis were used to run models that required 82 GB of RAM. Using the direct solver, the models take 45-60 minutes to run. It has also been possible to model a lymph node with spatial resolution of 1.38 µm and 2 x 107 elements. This required 110 GB of RAM and took 1 day and three hours to converge using an iterative solver, which without Iridis would have been achievable.

Acknowledgements: We gratefully acknowledge the assistance of Jürgen Mayer (EMBL/CRG Systems Biology Research Unit, Centre for Genomic Regulation (CRG) and UPF, Dr. Aiguader 88, 08003, Barcelona, Spain) who provided the SPIM lymph node images.

References: [1] T. H. Adair and A. C. Guyton. Modification of lymph by lymph nodes. III. Effect of increased lymph hydrostatic pressure. Am J Physiol. 1985, 249:H777-82.


Life sciences simulation: Biomathematics

Physical Systems and Engineering simulation: Biomechanics, CFD

Algorithms and computational methods: Finite elements

Visualisation and data handling methods: Voxel imaging

Simulation software: COMSOL

Visualisation and data handling software: Avizo, ImageJ/Fiji

Programming languages and libraries: Matlab

Computational platforms: Iridis

Transdisciplinary tags: IfLS