[ENH]: In-built support for spearman correlation for FunctionalConnectivity markers

[synchon]

Which marker do you want to include?

Currently it does not work out of the box, but some people like it and its a small enough thing to expect it out of the box and not really worth it have users implement their own markers for this.

Is there any publication or available code?

just rank time series and get pearson correlation

Do you have a sample code that implements this outside of junifer?

I can take a shot at implementing this if everyone agrees we should have it.

Anything else to say?

No response

[fraimondo]

This also relates to #333.

We need to somehow address the issue that nilearn does not accept any other "method" than these ones here: https://github.com/nilearn/nilearn/blob/4f4730163097457cf9ddb5674ffd158ee8fa822e/nilearn/connectome/connectivity_matrices.py#L497

[synchon]

In my opinion, we might need a custom JuniferConnectivityMeasure akin to JuniferNiftiSpheresMasker, which would allow this.

[LeSasse]

In my opinion, we might need a custom JuniferConnectivityMeasure akin to JuniferNiftiSpheresMasker, which would allow this.

In my opinion, this is reasonable, and would also allow us to set the default to EmpiricalCovariance which is what we use mainly anyways.

[synchon]

In my opinion, we might need a custom JuniferConnectivityMeasure akin to JuniferNiftiSpheresMasker, which would allow this.

In my opinion, this is reasonable, and would also allow us to set the default to EmpiricalCovariance which is what we use mainly anyways.

We definitely will have more upsides than downsides from what I understand. I'll get the skeleton ready if @juaml/junifer-core agrees on this and then you can get this in while I get #333 in?

[LeSasse]

In my opinion, we might need a custom JuniferConnectivityMeasure akin to JuniferNiftiSpheresMasker, which would allow this.

In my opinion, this is reasonable, and would also allow us to set the default to EmpiricalCovariance which is what we use mainly anyways.

We definitely will have more upsides than downsides from what I understand. I'll get the skeleton ready if @juaml/junifer-core agrees on this and then you can get this in while I get #333 in?

sounds good! ping me once i can start getting active

[synchon]

In my opinion, we might need a custom JuniferConnectivityMeasure akin to JuniferNiftiSpheresMasker, which would allow this.

In my opinion, this is reasonable, and would also allow us to set the default to EmpiricalCovariance which is what we use mainly anyways.

We definitely will have more upsides than downsides from what I understand. I'll get the skeleton ready if @juaml/junifer-core agrees on this and then you can get this in while I get #333 in?

sounds good! ping me once i can start getting active

@LeSasse You should be able to easily implement it by tweaking https://github.com/juaml/junifer/blob/main/junifer/external/nilearn/junifer_connectivity_measure.py

[LeSasse]

We need to define the value for kind, i guess, so maybe something like:

kind : {"covariance", "correlation", "partial correlation", "tangent", "precision"}

can turn into:

kind : {"covariance", "spearman", "pearson",  "partial correlation", "tangent", "precision"}

What do you think

[synchon]

We need to define the value for kind, i guess, so maybe something like:

kind : {"covariance", "correlation", "partial correlation", "tangent", "precision"}

can turn into:

kind : {"covariance", "spearman", "pearson",  "partial correlation", "tangent", "precision"}

What do you think

I'd actually keep "correlation" as is so that we're 100% compatible with nilearn but make the Spearman or any other correlation metric like so: "<name> correlation", so Spearman can become: "spearman correlation".

[LeSasse]

sounds good, then the list for implementing the spearman correlation will be:

kind : {"covariance", "spearman correlation", "correlation",  "partial correlation", "tangent", "precision"}

[synchon]

sounds good, then the list for implementing the spearman correlation will be:

kind : {"covariance", "spearman correlation", "correlation",  "partial correlation", "tangent", "precision"}

Exactly! :D

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.56%. Comparing base (675ccd0) to head (cc589c3).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #335      +/-   ##
==========================================
- Coverage   88.56%   88.56%   -0.01%     
==========================================
  Files         115      115              
  Lines        5178     5175       -3     
  Branches     1034     1032       -2     
==========================================
- Hits         4586     4583       -3     
  Misses        426      426              
  Partials      166      166              

Flag	Coverage Δ
docs	100.00% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files	Coverage Δ
...r/external/nilearn/junifer_connectivity_measure.py	92.30% <ø> (-0.22%)	⬇️

[github-actions]

PR Preview Action v1.4.7
Preview removed because the pull request was closed.
2024-08-06 12:51 UTC