Table Of Contents

Previous topic

DICOM importing

Next topic

Tissue class segmentation

MRI Quality Assessment

Structural MRI quality assessment

Visual inspection

The first time you look at an MRI image, you have very little with which to compare it, and so it’s unlikely you will have a good sense of whether it’s normal or not. Over time, you’ll build up a sense of what is normal and not, and you’ll become more accurate about detecting anomalies based on visual inspections. The only way to develop this experience is (not surprisingly) to make sure to always examine your data. However, even at the outset, there are some obvious errors that you can detect from visual inspection, including things like:

  • Partial-brain coverage
  • Large areas of signal dropout (e.g., due to a hairpin)
  • The wrong type of scan

I will include some examples of various artifacts here as I collect them.

The easiest way to view an image in SPM is to use the Display button (see Using SPM’s Display tool), or for multiple images, SPM’s CheckReg utility (see Using SPM’s CheckReg Tool).

Automated inspection

Although many artifacts are apparent after visual inspection, there are also many that aren’t. Thus, it’s also a good idea to use some sort of automated script to assess the quality of your images. A basic, but robust, method is to look for outliers in a population of subjects. Even using global values (i.e., a single value per image), this can often identify outliers based on initial intensity or segmented values (as problems in the initial image quality will also influence the quality of the segmentation).

After identifying outliers, you may want to exclude these outright, or take a closer look and see whether an error in processing may be rectified.