Atlas of Nasal Flows


  1. Introduction
  2. Model Extraction and Clean-up
  3. Boundary Conditions
  4. Simulation setup
  5. References


The Importance of the Human Nose

  • The importance of the human nose, as a vital part of the upper respiratory
    tract, is well known and well documented.
  • The nose has a complex anatomical structure:
    • internal soft tissue forms the nasal cavity
  • Such a structure makes it extremely difficult to generalise and understand the
    exact mechanism of nasal breathing.
  • A detailed insight into mechanism of nasal airflow and the consequential
    phenomena would contribute to the comprehension and identification of a
    variety of nasal conditions and selection of an appropriate medical treatment:
    • better understanding of recirculation in the nasal cavity which may
      lead to the persistence of bacterial infections,
    • better understanding of the function of sinuses and the treatment of sinusitis,
    • better understanding of the effects of a deviated septum on normal breathing,
    • better understanding of the effects of rhinoplasty.

The Atlas of Nasal Flows

  • The project investigates the idea of creating an atlas of healthy nasal airflows
    using Computational Fluid Dynamics (CFD):
    • The pattern of nasal airflow can be determined by examining a sufficient
      number of anatomies, in order to capture the significant
      characteristics of the flow.
    • A challenge of the simulation process is the description of the geometry
      and definition of boundary conditions.
  • Several modelling and meshing applications were used in conjunction
    with the open-source simulation library, foam-extend, to create a
    computational mesh, based on computational tomography (CT) scans.

The Workflow

  • A simplified version of the workflow used in project may be summarised as
    follows, in six major steps:
  1. Extracting a 3D CAD model of a nasal cavity from a CT scan.
  2. Cleaning up the 3D model to obtain a geometry suitable for the meshing algorithm.
  3. Creating an unstructured computational mesh using an automatic tool (cfMesh).
  4. Setting up boundary conditions for a transient, incompressible, turbulent flow simulation.
  5. Running the simulation with an implicitly coupled pressure-velocity solver.
  6. Processing the results: visualisation and analysis.

Model Extraction and Clean-up

  • The process of obtaining a 3D model consists of two main stages:

    • segmentation - the process of delineating structures of interest
    • volume rendering - process of visualizing volumes as 3D objects
  • The steps provided here, serve as general guidelines only.

  • Starting from a CT scan of the patient’s head, a 3D CAD model was created
    using 3D Slicer:

    • consecutive 2D images are combined into a 3D surface based on the colour
      saturation to distinguish empty space in the geometry from nasal tissue.
    • an example of the 3D model extraction process:
  • Extraction of the region of interest (nasal cavity) and division into
    separate surfaces (needed for defining boundary conditions) was done using
    Blender and MeshLab:

    • any changes to the geometry to simulate virtual surgery can easily be
      done during this step of the model creation procedure using Blender
    • examples of geometry clean-up in Blender (left) and MeshLab (right):
    Blender MeshLab
  • The spatial discretisation (creation computational mesh) was done in an
    unstructured manner, using cfMesh:

    • the result is an unstructured mesh consisting of 4 407 571 cells:
Blender Meshalb

Boundary Conditions

  • The boundary conditions are defined by using the experimental data
    obtained from literature.
  • The airflow was assumed to be incompressible and turbulent:the k − Ω SST
    RANS turbulence model with wall functions was used.
  • Because of the unknown flow conditions at the inlet point of the nasal cavity,
    a spherical boundary was added which corresponds to atmospheric conditions.
  • The beginning of the trachea was set as the outlet, with time-varying values
    of velocity set (taken from experimental data).
  • The boundary conditions for all variables are shown in the table below:

  • Final definition of boundary patches used:

Simulation setup

  • The coupling of pressure and velocity was solved using a transient,
    implicitly-coupled solver:
    • the pressure and velocity fields are solved in a single block-matrix
      which enables the implicit treatment of the pressure gradient in the
      momentum equation, and velocity divergence in the pressure equation.

    • we benefit from a more stable and faster simulation.

    • Three conditions occurring during nasal breathing were simulated:

      • normal inhalation lasting 2 seconds,
      • normal exhalation lasting 3 seconds,
      • quick inhalation ("sniff") lasting 0.5 seconds.

Normal inhalation

  • Case settings:
    • Normal inhalation lasting 2 seconds
    • The inflow segment of the breathing curve was used to calculate the
      inflow air velocity during inhalation:


  • A selection of the simulation results is presented next.

  • For the case of normal inhalation, distinct flow features can be observed:

    • Looking at the cross-section of the nasal cavity at several locations and
      time-steps, it can be seen that the largest velocity magnitude develops
      along the nasal septum which separates the nasal canals (7.3 m/s at the
      nasal valve region).
    • In the peripheral part of the nasal cavity (meatus region), the velocity
      is significantly smaller (3.6 m/s).
    • Weak airflow can be noticed in the sinuses (0.04 m/s) during the second
      half of the inhalation when the velocity starts to decrease in the first
      section of the nasal cavity.
  • For the particular patient, there is more obstruction in the right side of the
    cavity, therefore the velocities are higher in the left side of the cavity.

  • Videos depicting streamlines (left) and TKE (right):

  • TKE field in both nostrils:

Normal exhalation


  • Conclusions similar to those for normal inhalation, may be concluded here with
    some simulation-specific phenomena further examined in the project

  • Videos depicting streamlines (left) and TKE (right):

  • TKE field in both nostrils:

"Sniff" - a quick inhalation

  • Case settings:
    • a very quick and forceful inhalation lasting 0.5 seconds
    • The outflow segment of the breathing curve (shown above) was used to calculate the
      outflow air velocity during exhalation.


  • Conclusions similar to those for normal inhalation, may be concluded here with
    some simulation-specific phenomena further examined in the project

  • Videos depicting streamlines (left) and TKE (right):

  • TKE field in both nostrils:


  • All references are listed in the original project documentation:
    • Balatinec, Luka. Atlas strujanja u nosnoj šupljini (Atlas of Nasal Flows), 2018, Rector's Award Project,
      Faculty of Mechanical Engineering and Naval Architecture, University of Zagreb