Skip to content

GPU-CUDA toolbox for fitting compartmental models to 4D medical dynamic volumes, based on Maximum-a-Posteriori Levemberg-Marquardt optimization of non-linear kinetic models implemented using pyCUDA and cuBLAS python interfaces.

License

Notifications You must be signed in to change notification settings

mscipio/gpuKMfit

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Compartmental models parallel GPU-Cuda fitting toolbox

GPU-CUDA toolbox for fitting compartmental models to 4D medical dynamic volumes.

Several models have been already implemented (1TC, 2TC, 2TCr) in the big group of non-linear compartmental models. More to come (also linear, hopefully, like Patlak and Logan).

The optimization is based on a Maximum-a-Posteriori version of the standard Levemberg-Marquardt algorithm for non linear least squares optimization. A couple of local spatial prior are already available (quadratic and Huber). We plan to add more options, like TV and above all some anatomy-related prior, as well.

It uses pyCUDA and cuBLAS python interfaces to parallelize the LM non-linear optimization algorithm. Next version will probably switch from pyCUDA to CuPY for CUDA interface in python, but we still need to evaluate (suggestions welcome!)

TODO (first of all, tiding up everything .. sorry for the mess!)

  • Add instruction to how to install the library
  • Add instruction of how to use the class
  • Add some ipython notebooks with examples

About

GPU-CUDA toolbox for fitting compartmental models to 4D medical dynamic volumes, based on Maximum-a-Posteriori Levemberg-Marquardt optimization of non-linear kinetic models implemented using pyCUDA and cuBLAS python interfaces.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published