ETH Logo PSI Logo

Dark field of lung microstructures from synchrotron micro CT to a table top setup

From monochromatic beam and single sizes to a lung model

\(B = v / v_0 = \exp (-\mu_B t)\)

Sample made of identical spheres of the same material \(\mu_B(r, E; n, f)\) studied in recent literature:

  • \(r\) radius of the spheres
  • \(E\) beam energy
  • \(n\) refractive index
  • \(f\) volume fraction occupied by the spheres

Literature links

we used the latest for which we have the code

Polychromatic extension

\[ \mu_B = \int_{\text{radii}}\int_{\text{energies}} \mu(r, E; n, f)\rho(r)s(E)\mathop{dr}\mathop{dE} \]

input description source
\(s(E)\) source spectrum simulation
\(n\) refractive index NIST database
\(f\) volume fraction occupied by spheres microtomography
\(\rho(r)\) distribution of sphere sizes microtomography

Consistency check: integrating over energies only

Estimating \(n\) and \(f\)

  • \(n\): NIST database with density matching beamline absorption at 12 keV
  • \(f\): pixel counting in the segmented image

From lung microstructures to sphere distributions

Smoke stitched

Technical challenges

  • datasets with 2540x2160x6000 pixels
  • fiji implementation has excessive requirements on both memory and processing time

Complete rewrite

  • C and C++ with python bindings
  • code (with Goran and Arttu)
  • eliminate empty spaces outside of the sample
  • ignore airways larger than 100 μm
  • process spheres from the largest to the smallest, drastically reducing cpu cycles
  • split volume into several subvolumes for a smaller memory footprint
  • used all hardware resources: gws-3, gws-4 and RA cluster

Polychromatic extension

\[ \mu_B = \int_{\text{radii}}\int_{\text{energies}} \mu(r, E; n, f)\rho(r)s(E)\mathop{dr}\mathop{dE} \]

input description source
\(s(E)\) source spectrum simulation
\(n\) refractive index NIST database
\(f\) volume fraction occupied by spheres microtomography
\(\rho(r)\) distribution of sphere sizes microtomography

Results

Thank you for your attention

Backup

Previous progress report

Microtomography and postprocessing

  • synchrotron microtomography
  • segmentation
Segmentation procedure
G. Lovrić et al., Automated computer-assisted quantitative analysis of intact murine lungs at the alveolar scale, in press 10.1371/journal.pone.0183979

Alveoli as spheres

  • fit spheres in the lung microstructures
  • plot diameter distribution
Plos thickness
Size pdf
G. Lovrić et al., Automated computer-assisted quantitative analysis of intact murine lungs at the alveolar scale, in press 10.1371/journal.pone.0183979

The final model

sum over the sphere size distribution times the dark field response for each spere size

ground truth

Size pdf

dark-field response

Full summary