A Matlab-based PET simulation and reconstruction framework with analytical modeling of system matrix, capable of incorporating various degrees of spatial resolution’s Point Spread Function (PSF) modeling.
Key Features
- Analytic simulation (forward projection) of PET data (with and without noise)
- Begin with your own image, or you can use our included NEMA NU-2 phantom generator
- Enables noisy and noise-free reconstructions.
- Statistical PET image reconstruction framework using 2D ordered-subset expectation maximization (OS-EM) algorithm
- Incorporate attenuation and normalization modeling and correction using a CT sinogram and detector normalization map, respectively.
- Integrated spatially invariant generalized PSF modeling (resolution modeling) to model true PSF, no-PSF, as well as under- and overestimated PSF in reconstruction
- Can vary a range of data acquisition and image reconstruction parameters
- Capable of simulating and reconstructing signal (tumor) absent, i.e. replacing the signal value with the background, as well as signal present, to study the effects of model parameters on the background in the exact same location as the tumor.
Reference Links
Github page (Source Code, Documentation)
Qurit Lab Page (Presentation, Technical Description)
Reference Paper: S. Ashrafinia, et al., “Generalized PSF modeling for optimized quantitative-task performance”, Phys. Med. Biol., vol. 62, pp. 5149-5179, 2017.
For any questions or feedback, please contact: arman.rahmim@ubc.ca
NMMItools 