Recommended platform

Hello everyone,

I’m embarking on a new role as an image processing specialist within a research group that’s keen on utilizing SPM12 for a sizable dataset of PET images, totaling around 150 scans. My background is in biomedical engineering, and I’ve been exploring various approaches to implement SPM12 analysis effectively. I’m eager to engage in discussions regarding the merits and drawbacks of different platforms and software options available. This is in addition to the tutorials provided within MATLAB itself, which I’m still in the process of examining. I’d greatly appreciate your insights and recommendations on where to begin.

The platforms I’ve identified that support SPM12, based on their documentation, are as follows:

Furthermore, I’ve come across other platforms that appear intriguing for both project sharing and potential development, such as:

In my exploration of Neurodesk, I’ve discovered that it allows me to employ SPM12 pipelines within a Colab Jupyter notebook. This is particularly interesting as it enables cloud-based work, offering advantages like reproducibility and shareability. It also allows me to leverage machine learning (via SKlearn) and artificial intelligence (using TensorFlow) algorithms and pipelines for making inferences. This aligns well with one of our research objectives.

Regarding the Clinica platform, I’ve found it to offer a straightforward implementation of SPM12, requiring minimal calibration knowledge. They primarily utilize the GLM and provide user-friendly visualization of results through their pipelines. Additionally, there’s ClinicaDL, a deep learning extension, for those interested in exploring this avenue.

I admit that I haven’t fully grasped the capabilities of Brainlife and NeuroHub. While both claim to support Jupyter notebooks and development features, I couldn’t readily locate them in my initial exploration. Also, there is no documentation that i could make SPM12 work on these platforms.

In conclusion, I believe that Neurodesk with Nipype could be a more straightforward and efficient solution for implementing SPM12 on this dataset. This is especially appealing since I can design workflows and then apply them to all PET images, avoiding manual processing, which appears to be the case with Clinica.

Your insights and recommendations on these platforms and any additional tips or experiences you can share would be immensely valuable to me as I embark on this exciting endeavor. Thank you!

Don’t have a good answer as I do not do PET related things.

if you restrict yourself to nipype VS clinica, I would still first try with clinica as it already has some pipelines ready to go on standardized data (BIDS) where using nipype would require you to build your pipeline from scratch.

also not sure how you can use SPM12 on cloud platform unless you “package” it in a container with the extra tooling you need (nipype, clinica…): maybe using neurodocker could help with this. but if you are going down that road then maybe try to convince the neurodocker folks to install clinica into their image.

Thanks @Remi-Gau, I was exploring the possibilities of deploying Neurodesk (a popular neuro container with spm12, Matlab, and nipype) alongside the installation of clinical via pip. Apparently, these work on my Colaboratory notebook, since I could interact with clinica on a jupyter cell and also have the dependencies installed (or I think they are, I am new on these container software thing). Any advice you could give me to understand if this is possible or it does not make sense what I do on the notebook, i would appreciate it.

Also, with Clinica, the workflow is via BIDS structure files. But at my institute, they gave me a base SPM12 protocol based on NifTI files that someone did whoever knows when, and also it implies the use of some Python scripts I don’t have. I don’t know what to do, I understand that nowadays BIDS is becoming the standardized format, and I think the PET dataset they are to give me is in this structure format. I should prefer to do the analysis on BIDS structure or over simply PET NifTI files?.

You may want to have a look at this thread:

Well, I’ll strive to organize all the data in a BIDS structure when it arrives. Additionally, what about using SPM12 in MATLAB and batch scripts? I recently attempted one of the tutorials from the manual, and although the interfaces seemed user-friendly, they also provided a batch script with all the commands required to complete the same tutorial through the interface (one function crash on the interface, so i could not finish it from the interface). Consequently, I’m curious about the advantages of using Clinica versus implementing batch scripts for SPM. I understand that Clinica offers a ready-to-go batch methodology, but how can I determine if it’s the right fit for my analysis?.

hard to say or give a more specific answer beyond “it depends”, but If you data is “fairly” typical and your preprocessing is “fairly” typical, then starting with a standardized pipeline can be a time saver

having a very good idea what analysis you want to run and going for a deep dive in the documentation of the software you are interested in usually go a long way to figure out, if it is a right fit for your use case.