Resampling MRI Data: Common Uses and Pitfalls

A brief overview of what resampling is, how to use it, and how to avoid common mistakes.

Table of contents:

0:00 Introduction
0:49 What resampling looks like (using AFNI for the demonstration)
1:27 Nearest Neighbor resampling animation
2:27 When is resampling used?
3:00 Reference vs. Source Image
3:34 Common resampling mistakes
4:11 How to resample in the major packages (AFNI 4:27; FSL 4:47; SPM 5:06)
5:29 Summary & Valediction

[CognitiveInsurance]

Feeling so supeiror to other grad students that keep making the same mistakes over and over, also I am sharing this video

[wishswiss]

What is the meaning of reference volume. It only takes the reference's voxel dimensions, not the image content, am i right? ..
The image came in sagittal, but the transverse plane I put in was

[mansouralawi]

What about standardizing the image dimensions along with the voxels for example:
1)- MRI image for a child (voxel size [0.8 0.8 0.8] and dimensions [176X230X230])
2)- MRI image for a adult (voxel size [1.2 0.8 0.8] and dimensions [191X240X232])
if I resample both to (voxel size [1 1 1] and dimensions [255X255X255]),
will the anatomical values or characteristics such as the volume of grey matters or anatomical dimension of the new sampled images defer ?
ps*(aiming to do TMS comparison between the two images)

[notboredpanda]

Thanks for the explanation. It is indeed useful.

[dijifakoya]

Hello,
My name is Diji and I am working on brain sMRI data and I have some questions but the answers I have found on the internet and other platforms are a bit cumbersome or too technical for newbies in MRI neuro-imaging analysis like me.
Let me start :
Do I need to resample sMRI images? The aim is to get the classification between two cohorts of neuro-images which are progressive and stable Mild Cognitive Impairment using MRI data.
Secondly, is there any written law to follow when it comes the steps involved in image preprocessing before running the images through a model?

[abelalez]

what if the functional image is coregistered to the anatomical image?, so the functional dimensions change.

[ernestamoore]

Hi, I just left you a message on GitHub, just found your video. I am trying at the moment to resample a 61x73x61, time series of fmri data to something smaller to work with deep learning which is computationally very expensive. I should end up with a smaller cube (maybe 30x30x30), but I don't know how to go about it. Once the data is analyzed, I should upsample it back to its original shape. Any advice would be welcome. Thanks.