Newbie Question - coding imaging density / signal measurement

General
1

Greetings friends,

I’m new to using LOINC and I’m still trying to find my around for the purposes of coding imaging features, like size, attenuation (CT) and signal drops (MRI). While I did find size measurement at least for CT (96916-2), I’ve gone through all synonyms I could think of for measuring density or attentuation. I found that for the context of adrenal masses (95558-3), which is fine, but I also need a code for general density measurement and so far I haven’t found a good way for “measuring signal intensity on MRI”.

I strongly suspect user error, but I’m coming up empty. Can someone point me in the right direction? It’d be greatly appreciated!

Many thanks,

Jan

1 Like
2

Hi Jan,

If I understand the question correctly, signal intensity in the context of an MRI refers to the degree of “lightness” or “Darkness” on an MRI image and may be represented in LOINC as “Amplitude” similar to an the amplitude of an EEG?

I am not seeing a concept that fits this in LOINC, so it may require a request for new content. Something like “Image amplitude on MRI” where MRI is the method and I believe the scale might potentially be ordinal with an answer list of

  • Strong Signal - White
  • Intermediate Signal - Grey
  • Weak Signal - Black

I am not certain if a single concept is sufficient, or if the rating scale changes depending on the MRI method (e.g. FLAIR, T2, etc.)

Did I hit the mark or am I off in left field someplace?

1 Like
3

Hey John,

thank you for chiming in! And hats off regarding your analysis! Maybe I can add a little detail:

Signal intensity is a somewhat arbitrary value that an MRI scanner assigns to a voxel and is the result of biological (namely proton density - in humans, mostly water) but also technical aspects, so the value will be different from scanner to scanner. It has no real unit of measure, though there are some sequences (usually called mapping) that aim at creating objective, comparable measurements between scanners.

These signal intensities are rendered on the screen we use for reporting by assigning greyscale values; since the human eye can only discern so many different shades, we assign the values that routinely span intensities in the multiple hundreds into greyscale buckets, similar to CT, where we e.g. assign any value below -100 as completely black, no matter if -100 or -1000. The distribution of these assignments, and therefore how they appear on the screen depends on what we*re interested in and can vary wildly, so the same intensity can have vastly different “screen brightness” that we frequently adjust when reporting, which we call “windowing”. For LOINC, we’re therefore practically looking at an interval scale for all intents and purposes.

Now… it wasn’t my intention to cause work nobody else seems to need. That said, I suspect it might be a reasonable introduction, for the following reason:

Currently, LOINC has a code for measuring CT density of adrenals. This is because this measurement allows us to characterize a very common lesion called an adenoma in a large number of cases, because a low density of below 10 HU is considered diagnostic of an adenoma, rather than a more suspect lesion.

On MRI, we do something similar; we can’t measure density, but we can measure fat signal. A lesion loosing signal intensity if we suppress lipid signals would thus be considered an adenoma, and indeed, current guidelines name the use of MRI for characterizing lesions this way (the ESE 2023 guidelines, for instance).

Therefore, I’d argue that based on the inclusion of adrenal density measurement alone, the inclusion of measuring MRI signal intensity makes sense. It unquestioningly makes sense to have a more general code for CT density measurement outside of the adrenals, if there isn’t one - i couldn’t find it, but I still suspect i just overlooked it.

Does this make things more clear?

4

Hi Jan,

Thank you, and glad I was in the same ballpark with my answer.

I recently authored concepts that represent hypervascular intensities on FLAIR MRIs in SNOMED CT so had to do a good bit of reading to understand the difference between FLAIR and T2 rendering.

I think it makes sense to request intensity concepts in LOINC. Just as important as requesting the concept is the answer list interpretation associated with the requested LOINC code. If the answer list value definition includes a valid grey scale color range that is based on a color standard (i.e. CMYK, RGB, or pantone color standard) then the answer list becomes both human readable and machine computable. The Upper and lower limits of the CMYK range, for example, are the defining attributes or essentially a reference range for the human readable text interpretation.

I need some input and thought regarding which is the right color standard for use in this situation:

  • Pantone is the color standard for printed material
  • CMYK is the color standard for full-color images, but I don’t know if that applies to greyscale
  • RGB is the color standard for digital representation like computer monitors

There are likely maps between the three standards as I know Hexidecimal color codes can be represented in RGB, but this needs a bit of thought and intentionality put into proper selection, so we fit all the use cases. At the very least, if there is a curated map between the color palette standards we can use RGB as the base standard and point to the map for conversion into rendered image color palettes since these start out as digital images.

I know everyone likes to name things in their own way, but the answer list names should optimally reflect the color name from the selected standard color palette. Since these will essentially be defining attributes of the answer list value for the LOINC concept, we want to respect and leverage the reference standard by adhering to their naming conventions.

It is a little more upfront work, but the long-term benefits to modeling the content this way will be extremely beneficial for clinical decision support and data analytics.

Just a couple of thoughts that I hope help.

John

1 Like