renelabounek
commented
2 years ago Step 1: Define and unify qualitative assessment which we request An experienced radiologist scored the image between 0 (worst) and 5 (best), based on:
- The diagnostic relevance of the image
- The ability to distinguish GM from WM (contrast)
- The sharpness of the GM/WM border (affected by spatial resolution and motion)
- Signal drop-out due to intravoxel dephasing
- Ghosting artifacts (patterns of overlapped tissue on the cord along the phase-encoding direction).
I am sceptical to the point "The diagnostic relevance of the image". All images acquired the same tissue in the same FOV. Therefore, all images should have the same diagnostic relevance. I do not see what difference should be expected here over submissions.
Step 2: Make an instructions how to score each requested property: The dynamic range 0-5 gives me a binary excelent (1) or terible (0) for each of 5 listed properties. Thus, the scorers should use decimal values for requested properties over submissions; e.g. ability to distinguish GM from WM is terrbile for submission1 (score: 0.0), moderate for submission2 (score: 0.5) and excellent for submission3 (score: 1.0).
Step 3: Which image/s will they assess? First scan of each submission? Or both scans and final score per submission will be an average over two scans? I see 13 submissions. Thus, scorers will assess 13 or 26 scans.
Step 4: Fill assessment table I have designed a google sheet which scorers can copy, fill and send us back via email. I assumed they will assess both scans per submission. Is the table clear?
Step 5: List of scorers and contacts to them I assume we will switch the list over email between us.
Anything other needs to be prepared to make the work for scorers as easy as possible?
jcohenadad
MarcellaLagana
It is Figure 4.1.8 of the book "Quantitative MRI of the Spinal Cord" (Editors Julien and Claudia), so we can cite it.
This will be taken care of by Marcella Lagana and @renelabounek