Implement a method to create a population-based MRI template. Given an input of several subjects’ MRI, create one standard template MRI for the population. The method will utilize the MRI registration framework available in DIPY.
Hello Sir, I am Roshni ,M.Tech. Research Scholar in IIT Ropar, interested in this project.
I have done segmentation(Lesions Segmentation), classification projects in Medical CT Scan Data, X-Ray Data, etc, libraries used are- Keras, Tensorflow and language used - Python.
Kindly guide me further.
Thanks for reaching out. Did you get a chance to put together some prototypes or approximate details about how you would approach the problem statement? It is okay if your approach is incorrect but if share some ideas around how you would approach the project idea then the mentors can guide you better. Also do share any questions/queries you might have about this project in general.
I am tagging the mentors of this project to help you out(@bramsh@ShreyasFadnavis@pjsjongsung ). But for them to help you out, please can I request you to read through the project idea thoroughly and ask pointed queries/questions? That will help the mentors answer your queries better. As I’ve mentioned in above comment too, the project idea already has few steps mentioned. Ask yourself how you would go about solving this problem and come up with a brief proposal. It is okay if your approach is incorrect, the mentors will help you correct your approach but you will have to take the initiative and come up with initial steps of solving the problem.
hi @arnab1896. Veer shah here, I had a query regarding this project, can we use deep learning to generate more accurate templates.
Hoping to hear back from you
Hi @arnab1896 I’m a first year grad student at the University of Michigan, Ann Arbor specializing in computer vision. This project is interesting and I’d like to contribute to it. I haven’t worked with MRI data yet but I’m a quick study and am able to pick up things quickly. Hope to contribute.
Based on the limited reading I have done, the steps involved are
Alignment of the various scans,
Some sort of averaging over the different MRIs over the voxels to develop a reference template
Use ANT and other techniques to finetune the template.
Over the course of the next few days I will be diving more into the details of the steps but hope this rough idea is correct.
A few questions I had:
There are different methods of obtaining MRIs which result in different images. For this project do we assume that all the samples come from the same machine or do we develop a method that can deal with MRI scans from different machines with different settings?
Similar to @Veershah26 question, is there scope for deep learning in this project? Just from experience, working with encodings of the various MRIs might lead to a more robust template.
I am very interested in being a part of this project. I am a third year undergraduate in Computer Science at the University of St Andrews. I don’t have a lot of experience with the suggested technologies; however, I am eager to learn and delve into this kind of work.
I am proficient with Python and have some experience using artificial intelligence for image classification. I have been thrown into the deep end countless times at University so I am confident that I will be able to quickly learn and understand anything new that is relevant to the project.
I have done some preliminary research and have produced some notes as well as a broad approach to the project. This can be found here: https://bit.ly/3vYF6p2
Please let me know your thoughts on my current approach if you get a spare moment.
We will be working on one modality either T1 MRIs or FA images generated from Diffusion MRI.
No fMRI or Diffusion MRI. @Gregor_Soutar
Data can come from different machines.
We don’t necessarily need deep learning here but if you have any suggestion you can add that into your proposal and explain why and how it would benefit the template creation. @Veershah26
Take a look at the guidelines on how to contribute to DIPY . Making a small enhancement/bugfix/documentation fix/etc to DIPY already before applying for the GSoC can help you get some idea how things would work during the GSoC. The fix does not need to be related to your proposal. We have and will continue adding some beginner-friendly issues in Github. You can see some of them here .