our research
TomoLab
TomoLab: software for tomographic vision
This project is a Python 3.x porting of our previous project Occiput.io, which is now no longer actively supported.
TomoLab is (or will be) a tomographic reconstruction software for PET, PET-MRI and SPECT in 2D, 3D (volumetric) and 4D (spatio-temporal) for Python 3.x.
The software provides high-speed reconstruction using Graphics Processing Units (GPU).
Note: an NVidia CUDA-compatible GPU is required.
TomoLab can be utilized with arbitrary scanner geometries. It can be utilized for abstract tomographic reconstruction experiments to develop new algorithms and explore new system geometries, or to connect to real-world scanners, providing production quality image reconstruction with standard (MLEM, OSEM, Ordinary Poisson OSEM) and advanced algorithms.
TomoLab implements algorithms for motion correction (direct motion estimation), kinetic imaging, multi-modal reconstruction, respiratory and cardiac gated imaging. The source code contains Jupyter notebooks with examples.
Installation
- Clone the official Github repo [https://github.com/TomographyLab/TomoLab] wherever you like, on your workstation.
- Open a terminal window and
cdto the main directory of the package (i.e. you should see the setup.py file from usingls) - [optional but recommended] Activate the virtual environment in which you plan to install
TomoLab. - Execute
pip install .(this will install the local version of the repo usingpip, as this package is not yet available on the online server of PyPI)
If you have troubles with any of these steps, please just open an Issue on Github and we will try to sort it out.
IMPORTANT
In order to correctly load and useTomoLabyou need to compileNiftyRec.
Please, follow the instructions provided in the repo to do that.
Once you are done compilingNiftyRec, you need to add it to your path, so thatTomoLabis able to find it when it's imported. There are many ways you can do this, here's my suggestion:
~/.bashrcin your favorite text editorexport LD_LIBRARY_PATH=/lib:${LD_LIBRARY_PATH}~/.bashrcfile
Current status of the porting of Occiput.io (python 2.7) to TomoLab (python 3.x)
Reorganization of the code
- integrating major (
Occiput.io‘s) dependencies within the mainTomoLabproject - switching to relative imports throughout the code
- consistently following PEP8 style rules
- choosing a naming convention for modules, classes and variable (in therm of Captialization, usage of underscores, and so on) and keeping it consistent
- integrating major (
Simulation
- Several synthetic phantoms ready to be generated. A set of routines allow to create complex geometries (which may also be combined together by addition or subtraction), specifying the desired size and shape.
- Prepared a documentation notebook to showcase synthetic phantom generation capabilities
- Python 3.x interfaces to projection and backprojection operation successfully built on top of NityRec low level (C++, CUDA) libraries.
PET reconstruction
- Static reconstruction using OSEM and MLEM
- Implementing basic smoothing prior for OSL-MAP-OSEM
- Dynamic Reconstruction
- Class for efficiently handling 2D+t reconstruction (for research purpose)
- Cyclic Scan Reconstruction, informed of motion information coming from MR vNAV data
MR reconstruction
- Everything still needs to be checked after moving from python 2 to python 3
SPECT reconstruction
- Everything still needs to be checked after moving from python 2 to python 3
CT reconstruction
- CT reconstruction is not yet available. Anyway, it should be straightforward to leverage PE ray-tracing system for CT reconstruction, in the next future.
Image registration
- [ ]
(PET and DCE-MRI) Kinetic modeling
- [ ]