Podrobnosti studentského projektu

Téma:Segmentace vláken v papíru z mikro CT dat
Katedra:Katedra kybernetiky
Vedoucí:prof. Dr. Ing. Jan Kybic
Vypsáno jako:Diplomová práce, Bakalářská práce, Semestrální projekt
Popis:The aim of the project is to create a method for automatic
segmentation of paper fibers from micro CT images.


1. Read the data and manual segmentations and visualize them.
2. Get acquainted with existing methods based on provided references.
3. Get acquainted with the PyTorch library, try some simple image
segmentation task using the UNet architecture.
4. Predict the fiber location in 2D using deep learning. I recommend
to formulate the task as a regression problem on the transformed
distance map (perhaps the Gaussian of the distance to the nearest
fiber center), similar to Naylor et al. or Mertanová.
5. Connect the fiber centers using a suitable algorithm (minimum path
methods, particle filtering, iterative global methods)
6. Evaluate the segmentation quantitatively on the paper fibers as
well as on other fiber datasets.
7. Consider existing deep learning approaches (e.g. DeepTract, Learn
to Track, Turetken et al. ) and suggest how to apply them on our data.
Literatura: Thomas Dietenbeck, François Varray, Jan Kybic, Olivier Basset and Christian Cachard. "Neuromuscular fiber segmentation through particle filtering and discrete optimization." SPIE Medical Imaging, vol. 9034, pp. 90340B, February 2014.

Learn to Track: Deep Learning for Tractography

Benou, Raviv: DeepTract: A Probabilistic Deep Learning Framework for White Matter Fiber
Tractography. 2018

Naylor: Segmentation of Nuclei in Histopathology Images by Deep Regression
of the Distance Map. IEEE TMI 2019.

Mertanová: "Cell segmentation in microscopy using a reference
modality". CTU diploma thesis, 2021

Turetken et al: Reconstructing Curvilinear Networks using
Path Classifiers and Integer Programming. IEEE PAMI 2016
DOI : 10.1109/Tpami.2016.2519025.
Za obsah zodpovídá: Petr Pošík