Skip to content

Discussing different methods of calculating Phase Amplitude Coupling

Notifications You must be signed in to change notification settings

MOH-Shafizadegan/In-search-of-true-PAC

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 
 
 

Repository files navigation

In-search-of-true-PAC

Discussing different methods of calculating Phase Amplitude Coupling

PAC Methods

  • CWT & MVL
  • Rid-rihaczek (old version, supposed to have implementation bug) & MVL
  • Rid-rihaczek (new version (neuroFreq), First tf-decomposition, then windowing) & MVL
  • Rid-rihaczek (old version, correctred versio) & MVL
  • Rid-rihaczek (new version (neuroFreq), First windowing, then tf-decomposition) & MVL
  • CWT & MI

Descripption

  • Synthesizing the signal:

    • The phase amplitude coupled signal has been generated following proposed method at the this article.
    • Generating a signal composed of:
      • Random gaussian nosie of 1 second
      • coupled signal one:
        • Amplitude frequency: 40 Hz
        • Phase frequency: 5 Hz
      • coupled signal two:
        • Amplitude frequency: 60 Hz
        • Phase frequency: 9 Hz
      • coupled signal one + additive noise
      • coupled signal two + additive noise
  • PAC Comodulogram

    • Generating comodulogram for each of the 5 sections of the signal
  • PAC dynamic

    • Generating PAC dynamic through all of the 5 second signal
    • Window length = 200 ms
    • Window shift step = 100 ms
  • Statistical test

    • Sample size: 100
    • Population 1:
      • Coupled signals:
        • Random Phase freq in range: [4, 7]
        • Random Amp freq in range: [38, 42]
    • Population 2:
      • Coupled signals:
        • Random Phase freq in range: [8, 11]
        • Random Amp freq in range: [55, 65]
    • Null hypothesis:
      • PAC in range [4,8] and [35, 45] is higher for second population

Conclusion

  • Each of these methods has their own pros and cons
  • Rid-rihaczek can decompose frequency more precicely than the CWT. The CWT will skew the freuqency axes
  • MVL method is biased with amplitude signal and power (Note the comodulograms)
  • MI requires signals long enough (more than 500 samples)
  • MVL may detect fake couplings too but works preciecly well with actual couplings
  • MVL doesn't require permutations test but using it may lead to better results espcially when we use CWT
  • Curerntly MVL is much faster than MI

Releases

No releases published

Packages

No packages published

Languages