To-do

[Bradley-Karat]

List of tasks to-do:

• Add in new models (DMIPY):
  • Ball and sticks (two and three)
  • Cylinder
  • Ball and racket
  • NODDI-Bingham
  • Multi-tissue NODDI (assumes different S0 for each compartment)
  • AxCaliber
  • IVIM
  • VERDICT
• Add in new models (DIPY):
  • DTI
  • DKI
• Add in another jupyter notebook linking uGUIDE and DiffSimGen:
  • i.e. what to do with your simulated data
• Attempt to learn and inject realistic noise into simulated signal from real data with GAN:
  • First fit NLLS to data to get realistic parameter distribution
  • Simulate ground-truth signal with the above parameters (no noise)
  • Train GAN to discriminate between real noisy signal and noiseless simulated signal (the generators job is to inject noise to make noiseless signal look like real signal)
  • Compare the GAN outputs to standard ways of injecting noise (SNR, noisemaps from dwidenoise, etc...)
• Build a toolbox for supervised model fitting based on data simulated from DiffSimGen: