makism / dyconnmap Goto Github PK

Hi
I am interested in the topoplot() function. But I do not find it available in the dyfunconnpackage.

Hi there,

Thanks for developing this package! I have already found some useful tools for static FC here, but I have been struggling to work out if it is possible to do any dynamic FC analyses for fMRI data? Following up on previous threads #11

If so, I'd be really grateful for some tips to get started. I'd like to start with generating some FC networks from sliding windows.

Many thanks,
Joff

It would be nice to support python3 in the near future.

A question of usage rather than a bug report.

Assume we have a 10sec EEG 14 channels data sampled at 128Hz. The objective is to evaluate the dynamic FC in a non-overlapping time windows of 2s. Given the sync bandwidth of interest is [1.0, 4.0].

Since, the lowest freq is 1.0Hz with 128Hz sampling rate, Cycle criterion should be equal to 2 (correct me if I am wrong).

Then, my question, what is the correct number for step value? My basic understanding tell me it should be equal to 5? Since, each of the 2sec window contain 256 sample, and moving the window for 5 times will cover all the 256*5 =1280 data points.

Say if it is correct step equal to 5. According to the line below, the total number of sliding windows as calculated from (n_samples - window_length) / step is equal to 205.

But this lead to another question, why the fcgs = np.zeros((windows, n_channels, n_channels)) having the shape of 205,14,14. Shouldt it be the size of 5,14,14?

In your first EEG tutorial, the eegmmidb EEG data is loaded with pyedflib. However, the EDF files in that project are not strictly complient, because they contain prefilter information in both the "standard" fields and EDF+ annotation fields. pyedflib chokes on them:

>>> import pyedflib
>>> pyedflib.EdfReader('eegmmidb/S001/S001R01.edf')

Traceback (most recent call last):
  File "pyedflib/_extensions/_pyedflib.pyx", line 146, in pyedflib._extensions._pyedflib.CyEdfReader.__init__
  File "pyedflib/_extensions/_pyedflib.pyx", line 207, in pyedflib._extensions._pyedflib.CyEdfReader.open
  File "pyedflib/_extensions/_pyedflib.pyx", line 179, in pyedflib._extensions._pyedflib.CyEdfReader.check_open_ok
OSError: eegmmidb/S001/S001R01.edf: the file is not EDF(+) or BDF(+) compliant (Prefilter)

I've send a pull request to pyedflib to address the issue, but how did you work around it when writing the tutorial?

Hi,

when using the latest Scipy v1.4.1, importing dyconnmap causes the following error:

---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-2-cc4061b0c448> in <module>
----> 1 import dyconnmap

~/dev/tf2/lib/python3.7/site-packages/dyconnmap/__init__.py in <module>
      5 """
      6 
----> 7 from .tvfcgs import tvfcg, tvfcg_ts, tvfcg_cfc, tvfcg_compute_windows
      8 from .sliding_window import sliding_window_indx, sliding_window
      9 from .analytic_signal import analytic_signal

~/dev/tf2/lib/python3.7/site-packages/dyconnmap/tvfcgs.py in <module>
     49 import numpy as np
     50 
---> 51 from .fc.estimator import Estimator
     52 
     53 

~/dev/tf2/lib/python3.7/site-packages/dyconnmap/fc/__init__.py in <module>
     14 from .cos import cos
     15 from .pec import pec
---> 16 from .glm import glm
     17 from .pac import PAC, pac
     18 from .mui import mutual_information

~/dev/tf2/lib/python3.7/site-packages/dyconnmap/fc/glm.py in <module>
     27 
     28 import numpy as np
---> 29 import statsmodels.api as sm
     30 
     31 

~/dev/tf2/lib/python3.7/site-packages/statsmodels/api.py in <module>
     14 from . import robust
     15 from .robust.robust_linear_model import RLM
---> 16 from .discrete.discrete_model import (Poisson, Logit, Probit,
     17                                       MNLogit, NegativeBinomial,
     18                                       GeneralizedPoisson,

~/dev/tf2/lib/python3.7/site-packages/statsmodels/discrete/discrete_model.py in <module>
     43 
     44 from statsmodels.base.l1_slsqp import fit_l1_slsqp
---> 45 from statsmodels.distributions import genpoisson_p
     46 
     47 try:

~/dev/tf2/lib/python3.7/site-packages/statsmodels/distributions/__init__.py in <module>
      1 from .empirical_distribution import ECDF, monotone_fn_inverter, StepFunction
----> 2 from .edgeworth import ExpandedNormal
      3 from .discrete import genpoisson_p, zipoisson, zigenpoisson, zinegbin

~/dev/tf2/lib/python3.7/site-packages/statsmodels/distributions/edgeworth.py in <module>
      5 import numpy as np
      6 from numpy.polynomial.hermite_e import HermiteE
----> 7 from scipy.misc import factorial
      8 from scipy.stats import rv_continuous
      9 import scipy.special as special

ImportError: cannot import name 'factorial' from 'scipy.misc' (/home/phait/dev/tf2/lib/python3.7/site-packages/scipy/misc/__init__.py)

An easy fix would be to change the import to:

from scipy.special import factorial

Cheers,
Daniele

I would like to ask, can ‘tvfcg_ts’ be used in the time series of fMRI? If so, is the parameter 'n_channels' equivalent to the ‘number of ROIs’ in the fMRI time series, and is 'n_samples' equivalent to 'slice' in the fMRI time series, and other parameters such as ‘fb’, 'fs', etc. how to set?

It came into my attention that it would be a nice idea if the parameters fs, fb (those related to filtering) were decoupled from the estimators themselves.

This way, one can use the estimators with, i.e. preprocessed fMRI time series.

In tutorial 5, the code attempts to load autism_fmri_ts.npy. Where can I find this file?

Dear Makism,

Hello.

First of all, thanks for the great tool.

After installation, I realized the test does not run correctly on my side. Could you have a look the output?
Is there anything I should modify?

slurm@master:/home2/data/dyconn/dyconnmap-master/tests$ ls
data test_fc.py
init.py test_graphs.py
pycache test_graphs_threshold.py
test_chronnectomics.py test_sliding_window.py
test_clustering.py test_ts.py
test_clustering_validity.py test_tvfcgs.py

slurm@master:/home2/data/dyconn/dyconnmap-master/tests$ nosetests -svd .

Ran 0 tests in 0.000s

OK

Yours sincerely
Meng

Environment: Python 3.7.2 (running in venv module), macOS 10.14.3

(env) $ pip3 install dyfunconn
Collecting dyfunconn
  Could not find a version that satisfies the requirement dyfunconn (from versions: )
No matching distribution found for dyfunconn

I see that the module is not published on PyPi, so you might just need to change the README to explicitly point to the git repo (if that is the desired op). When searching for it on PyPi the only result that comes up is Spektral.

Dear Makism or other users,
first of all, thanks so much for your contribution in dynamic connectome mapping, which I learned a lot.
But now I am confused of the usage of surrogate analysis, which, from my understanding, is aimed to identify the "true" connections between nodes.
Here is the procedure I understand:
firstly, I contructed a 4-D dynamic functional connectivity matrix (coupling modes × temporal segments × ROIs × ROIs) for each participant.
Secondly, the time serials from corresponding frequency band, two ROIs were extracted for surrogate analysis.
Thirdly, the dynamic functional connectivity values were compared with the 1000 surrogate values and then corrected by FDR.
Is that corrected?

Meanwhile, I have some confused points:

1. what's the meaning of the output "p_val, corr_surr, surrogates, r_value"?
2. the surrogate analysis was conducted in individual level or group level?

I would appreciate a lot if any one could answer my questions.

Hi,

I notice that you changed the project name and I was wondering if it means that significant changes are on the horizon...

I have a project that depends on dyconnmap and I just wanted to know if I should keep an eye out.

Thanks

r_diff evaluates as 1 for all elements, which yields a zero length array.

we should handle this as an edge case

Create a BibTeX file containing all the referenced citations within the module.