ah, gotcha - many thanks, Henry, much appreciated

g

On Fri, Apr 12, 2024 at 11:19 AM Tregidgo, Henry <h.tregidgo@ucl.ac.uk> wrote:

The Bayesian code is in the development version of Freesurfer yes.

 

At the moment this is the original compiled Matlab code used in development of the algorithm. Running this requires a matlab runtime similar to the original segmentThalamicNuclei.sh script from Freesurfer 7.0. We have someone working to integrate it into the same python framework as SAMSEG and segment_subfields but we are limited by available developer time.  

 

BW,

Henry

 

From: Georgios Argyropoulos <gargyropoulos@gmail.com>
Date: Friday, 12 April 2024 at 11:04
To: Tregidgo, Henry <h.tregidgo@ucl.ac.uk>
Cc: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>
Subject: Re: [Freesurfer] Question on using thalseg_1.1.h5

many thanks, Henry, that's super helpful. Btw, I came across the segmentThalamicNuclei_DTI.sh here (https://surfer.nmr.mgh.harvard.edu/pub/dist/freesurfer/dev_binaries/centos7_x86_64/?C=M;O=D). am I right in thinking that your Bayesian module is now integrated into the Python framework? No worries if not, obvs, can't wait to run it, I guess ! :)

 

many thanks again

 

g

 

On Fri, Apr 12, 2024 at 10:59 AM Tregidgo, Henry <h.tregidgo@ucl.ac.uk> wrote:

No problem Georgios,

 

It’s easily done. I’ll check if we can streamline the patching process.

 

I’ve written a bit of an explanation below in case you or anyone else on the freesurfer list comes across a problem like this again.

 

Best wishes,

Henry

 

===========================

 

Explanation:

Many of the network-based tools in freesurfer are distributed as python scripts which build a tensorflow or pytorch network, load the model weights from $FREESURFER_HOME/models and push the data through. As changing the combination of python package versions can have unexpected effects, freesurfer comes with its own python environment “fspython”. For these python-based tools the files in $FREESURFER_HOME/bin are shell scripts that run some checks then pass the python scripts in $FREESURFER_HOME/python/scritps and any arguments to the fspython environment.

 

In this case replacing the file in the bin directory with the python code likely meant it was running in your local python environment using different package versions. A clue to this is the deprecation warning mentioning numpy version 1.25 as the version in fspython is 1.24.3

 

From: Georgios Argyropoulos <gargyropoulos@gmail.com>
Date: Thursday, 11 April 2024 at 17:24
To: freesurfer@nmr.mgh.harvard.edu <freesurfer@nmr.mgh.harvard.edu>, Tregidgo, Henry <h.tregidgo@ucl.ac.uk>
Subject: Re: [Freesurfer] Question on using thalseg_1.1.h5

⚠ Caution: External sender

 

Many thanks and apologies, Henry - I figured it out... I stupidly replaced your improved segmentation script in both the scripts and the bin directory instead of only replacing the former.

 

many thanks for your time, much appreciated

 

g

 

--

Georgios P. D. Argyropoulos

PhD, CPsychol, FHEA

Division of Psychology, Faculty of Natural Sciences; University of Stirling; FK9 4LA Scotland UK

Personal Website

--

Georgios P. D. Argyropoulos

PhD, CPsychol, FHEA

Division of Psychology, Faculty of Natural Sciences; University of Stirling; FK9 4LA Scotland UK

Personal Website