Skip to content

Latest commit

 

History

History
109 lines (90 loc) · 4.06 KB

README.md

File metadata and controls

109 lines (90 loc) · 4.06 KB
Raw

CIL-HTC2022-Algo2

Authors:

  • Gemma Fardell (STFC), United Kingdom
  • Jakob Sauer Jørgensen (DTU), Denmark
  • Laura Murgatroyd (STFC), United Kingdom
  • Evangelos Papoutsellis (STFC, Finden), United Kingdom
  • Edoardo Pasca (STFC), United Kingdom

Addresses

STFC: United Kingdom Research and Innovation Scientific Computing Department of Science Technology Facilities Council Rutherford Appleton Laboratory Harwell Campus Didcot OX11 0QX

DTU: Technical University of Denmark, Department of Applied Mathematics and Computer Science Richard Petersens Plads Building 324 2800 Kgs. Lyngby Denmark

Finden: Building R71, Rutherford Appleton Laboratory, Harwell, Oxford, OX11 0QX

Description of the algorithm

This is an entry for the HTC2022 competition. The algorithm in algo.py is developed using CIL, a toolkit for tomographic imaging and optimisation. The main steps of the algorithms are:

  1. Pre-processing: renormalisation, single material beam hardening correction, zero padding
  2. generation of pixelwise lower and upper bound circular masks (mask is fitted to the provided data)
  3. Regularised iterative reconstruction algorithm using tools from CIL: Least Squares with both isotropic and 1-dimensional anisotropic TV regularisation, where orientation of projections is pre-aligned with one coordinate axis to get 1-dimensional anisotropic to encourage edges in the difficult direction orthogonal to the central projection angle.
  4. Post-processing: segmentation of the reconstruction with multi-Otsu threshold

Installation instructions

Installation instructions, including any requirements.

conda env create --file environment.yml

Usage instructions.

conda activate htc-22-cil-algo2
python main.py path/to/input/files path/to/output/files 3

Examples

Examples of reconstructing the example datasets, where we have limited the angles to 90, 60 and 30 degree ranges. The 'ref' column is the given segmented result from the full dataset.

Ref 90 60 30
ta
tb
tc
td

Scores for each sample and angle:

90 60 30
ta 0.988 0.963 0.916
tb 0.933 0.892 0.785
tc 0.956 0.918 0.830
td 0.967 0.960 0.940

Repository content

  • utils.py
  • algo.py
  • main.py
  • environment.yml
  • README.md
  • recalc_score.py
  • test_data
    • htc2022_ta_full_recon_fbp_seg.png
    • htc2022_ta_sparse_example.mat
  • results
    • htc2022_ta_sparse_example.png
    • AR30
      • htc2022_ta_full.png
      • htc2022_tb_full.png
      • htc2022_tc_full.png
      • htc2022_td_full.png
    • AR60
      • htc2022_ta_full.png
      • htc2022_tb_full.png
      • htc2022_tc_full.png
      • htc2022_td_full.png
    • AR90
      • htc2022_ta_full.png
      • htc2022_tb_full.png
      • htc2022_tc_full.png
      • htc2022_td_full.png

License

All files in the repository come with the Apache-v2.0 license unless differently specified.