search

MaastrichtU-IDS / dsri-documentation

📖 Documentation for the Data Science Research Infrastructure at Maastricht University
https://dsri.maastrichtuniversity.nl
MIT License
21 stars 8 forks source link

Request deployment of Neuro and imaging applications #21

Closed vemonet closed 2 years ago

vemonet commented 3 years ago

Describe the application

Deploy images from the NeuroDocker Project: https://miykael.github.io/nipype_tutorial/notebooks/introduction_neurodocker.html

neurodebian

Matlab packages: neuroelf, spm12, with the matlab UI

Application Git repository URL

https://miykael.github.io/nipype_tutorial/notebooks/introduction_neurodocker.html

FSLeyes UI:

https://hub.docker.com/r/kaczmarj/fsleyes

Notable licensing issues

Matlab

@fdemarti

fdemarti commented 3 years ago

other useful packages: MRtrix3 (for diffusion imaging) dcm2niix (this should go in any FSL docker if possible)

also note that the matlab packages SPM12 and neuroelf are useful ALSO if the UI is not available.

vemonet commented 3 years ago

@fdemarti we are in discussion with UM to access Matlab licenses but the number of licenses for Matlab at UM are limited

But researchers should not be using Matlab if they are doing Research or Education funded by public organizations

A part if they have no other choices (which is almost never the case). One of the main goal of research is to produce valuable and reproducible analysis.

When a researcher publish an analysis using a closed software like Matlab, he is publishing something that cannot be reused by other scientists, because they would need to pay for Matlab license and support (if matlab still exist the year you want to use it!)

Unfortunately, what researchers working with Matlab are doing is to promote a Matlab monopoly in science for the years to come (since scientists will be dependant on matlab later, even if there are better solutions)

I would recommend to any scientist doing public funded science to not use matlab, and use Open Source solutions like R, Python, or Julia

We can already see the changes within UM: before, Maastro research was almost 100% matlab, now most people switched to python

Open sources solutions will bring a enormous ecosystem of packages that matlab will never be able to reach

Here is some feedback from someone who worked with a different research softwares such as matlab:

I’ve had to deal with a lot of MATLAB sh** during my maastro time you pay to compile any kind of application, and you have to compile anything on the platform for which you intend it the smallest variation in environment from the system where you compile to the system you run on will give weird errors in obscure libraries, and then it takes you a long time to figure out that is the issue

The worst is that by using Matlab researchers are making the public to pay for this bad product, and they force the next generation of scientists to pay and use this bad product: please, for the sake of science, let's stop using Matlab.

Tutorials to switch from matlab to Python can be easily found here: https://realpython.com/matlab-vs-python/

vemonet commented 3 years ago

We have published some documentation to build and deploy NeuroDocker images: https://maastrichtu-ids.github.io/dsri-documentation/docs/neuroscience

You can also now easily deploy matlab from a template, see more informations here: https://maastrichtu-ids.github.io/dsri-documentation/docs/deploy-matlab

vemonet commented 2 years ago

You can find documentation to deploy popular Neuroscience software (e.g. FSL) here: https://maastrichtu-ids.github.io/dsri-documentation/docs/neuroscience

And some example Dockerfile to build a workspace with FSL using Neurodocker (for CPU or GPU) here: https://github.com/MaastrichtU-IDS/jupyterlab