search

odlgroup / odl

Operator Discretization Library https://odlgroup.github.io/odl/
Mozilla Public License 2.0
368 stars 105 forks source link

DBT Geometry and FBP Reconstruction ? #1597

Open mozanunal opened 3 years ago

mozanunal commented 3 years ago

Hello, First of all, thank you for this great project. We are a research team from Istanbul. We would like to use ODL project for Digital Breast Tomosynthesis (DBT) reconstruction.

We have a look at the geometry documentation at https://odlgroup.github.io/odl/odl.tomo.geometry.html However, we could not figure out how we should arrange geometry for Digital Breast Tomosynthesis. Do you have any example code or getting started guide for DBT reconstruction?

Thanks in advance.

adler-j commented 3 years ago

I must admit I'm a bit ignorant about DBT, do you have a more specific description of the geometry you want? My understanding is that you have a fixed detector and only move the source?

ozanoktem commented 3 years ago

I also think the detector collimators are slanted in tomosynthesis and the source moves along a linear trajectory. At least hat was the case with the X-Ray detector for tomosynthesis mammography by XCounter.

mozanunal commented 3 years ago

The devices which we have used so far are Hologic, Siemens, and GE. The detector collimators might vary from one device to another however the source moves in a linear directory for all devices as you mentioned.

adler-j commented 3 years ago

I don't think we have such geometries hardcoded into ODL right now since we haven't worked on it. To do it "nicely" you'd need to add a new DivergentBeamGeometry. ConeBeamGeometry is an example and while the implementation looks rather lengthy it's almost all documentation. If you want to implement one I can help out, otherwise you might need to wait a but until someone (me perhaps) does it.

mozanunal commented 3 years ago

I have another project on my focus now. I guess my project can take around 1 month. After that, I can focus on this one and try to implement it. Thanks for your kind interests and answers @adler-j @ozanoktem.

Another issue, I actually found an article about DBT reconstruction and I predicted that it is implemented with ODL toolbox (Since it is based on learned primal-dual). Thats why I want to contact you. I guess you are also in this publication if I am right? Maybe I should contact the first author.

https://arxiv.org/pdf/1808.04640.pdf