thijsdhollander
commented
3 years ago Hi,
I mean this honestly constructively, so it'd be nice to keep that spirit. This is not to attack anyone or similar.
I've recently checked this in detail, since I fully agree this seems to have caused some confusion over time. That said, the original 2 works that introduced the FBA framework (the CFE paper and the FBA / FD-FC paper) have quite explicitly used "apparent fibre density" or "apparent FD" (or at times a shortening to just "FD" indeed) to refer to the fixel-wise metric throughout. Searching for the word "apparent" in these sources highlights most of the relevant examples of this usage. There's a few paragraphs suggesting interchangeable language, but to bring up what I think are relevant statements:
- The abstract of the CFE paper, arguably the "first" occurrence of a statement describing what is analysed in the first actual FBA (the MND study / population): "We then demonstrate the proposed method by comparing apparent fibre density between motor neurone disease (MND) patients with control subjects."
- The results, e.g. the figure depicting them with caption: "Fixel-based analysis results demonstrating a significant decrease (FWE-corrected p < 0.05) in apparent fibre density (AFD) in motor neurone disease (MND) patients compared to healthy controls."
...and yes, anything in that paper of course, because "FD" only popped up in the other FBA paper. I have had discussions with David at that time even when he was trying to establish some terminology, and that exercise did indeed come about because of the introduction of FC. "FD" was chosen not to drop the notion of "apparent", but because we all know long complicated names don't adopt that well ("HMOA" doesn't read that well arguably, and its expansion can be a bit overwhelming I can imagine). But I think most importantly, this notion and language were not dropped in that (the FBA or "FD-FC") paper. The FD term is introduced to capture a potentially broader range of other dMRI models that "an" FD metric could be pulled from (and that could still lead to an FD-FC-FDC type of analysis). The methods section then clarifies what is used specifically when the fixel-wise metric is derived from the SD method:
- (in 3.2 "Fibre density"): "As listed in the Background section Fibre density (FD), there are several DW models that aspire to estimate quantitative measures related to fibre density. Any fixel-based measure can be employed in a fixel-based analysis; however, in this work we used apparent fibre density (AFD) (Raffelt et al., 2012b), a quantitative measure related to FD derived from FOD images computed from single-shell DWI."
So I agree that the FD term was introduced to contrast with FC indeed, and complement both or combine into the FDC notion; though not because of the dixel vs fixel difference. From the above wording, you could say something like "AFD happens to be the (fixel-wise) FD metric that we use when it originates from SD". So very broadly, there's an aspect that doesn't make them interchangeable at all times, because you might have an FBA study where "the FD metric" comes from something else, i.e. "is not AFD". But in most, if not all, current FBA studies out there, "the FD metric" is AFD. Agreed that doesn't make them linguistically interchangeable, but in a lot of sentences (but maybe not all possible sentences) you could find in such studies, you could use the one or the other and still be correct. Ultimately, the FBA paper also has a nomenclature section of course. It first established the point again that many names have been used before for "similar" (but not the same) concepts, and they all have technical problems when you're sufficiently strict about things. It then goes on to state:
- "Furthermore, while some multi-compartment DWI models explicitly ensure the volume fractions of all compartments sum to unity (e.g. Assaf and Basser, 2005), the apparent fibre density measure used in the present study is proportional to the measured DW signal, and hence not explicitly a volume fraction (Raffelt et al., 2012b)."
Referring back to what I think would be some of the most important reasons to say "apparent": it's not a true density, it's not a (volume or other) fraction, etc.. Again it also says that the fixel-wise (FD) metric used in the work itself is apparent fibre density. And then it makes the explicit point about why then "FD" is introduced as a term in the paper:
- "In this work we opted for the term ‘fibre density’ (FD) in part because it is already common in the literature, and also since it is most easily interpretable by non-technical audiences."
I.e. not because fixels are involved, but more as an opportunity to allow use of a more "simple" term. On the side, I do like that this kind of disclaimer (of sorts) is there, because we know that e.g. NODDI's "neurite density" has received backlash for maybe not coming with enough disclaimers. That just aside. But ultimately, scrolling back to the methods section for definitions, there's:
- "Briefly, FOD lobes are first segmented based on the peaks and troughs of the FOD, and the apparent FD of each lobe is calculated by non-parametric numerical integration using a dense sampling of the FOD over a hemisphere (Smith et al., 2013)."
I really like that this crucial sentence used "FD" in abbreviated form, but added "apparent" written in full before it. I also started using this frequently in writing to occasionally remind people of the "disclaimer" that the "apparent" term embodies. "FD" is easier to say, write, and "not confuse" some audiences unnecessarily, but "apparent FD / AFD / ..." is a good reminder or disclaimer to add nuance when it's needed (and only when "FD" comes from SD of course).
Also, somewhat interestingly, there's even a reverse kind of statement in the paper:
- "However, as demonstrated in our previous work (Raffelt et al., 2012b), fibre density and cross-section information can be combined to enable a more complete investigation of white matter."
...referring to the dixel-wise previous work, yet using "fibre density". Not to look for too much in that statement, but it's a similar lighter way of just saying "fibre density", even though the work did use AFD in particular. But similarly, you could go and use "FD" and even "FC" back in the dixel-wise work without issue, I think. (technically, that is; I'm not encouraging unnecessary confusion of course)
So:
But I also think that attempting to drop "AFD" entirely due to the change in domain from dixel-wise to fixel-wise is maybe not wholly warranted.
I think this is important. But because of this, I think it might be useful to separate the "dixels versus fixels" entirely from the "AFD versus FD" questions. They each have their own language / terminology to distinguish after all, right? Conflating them, I think, makes things less clear and might increase confusion. None of the assumptions or context / nuance that make "AFD" an "apparent" FD are gone or should be forgotten when moving to a fixel domain. Maybe:
The change from "AFD" to "FD" happened kind of "silently"
...what I mean is this wasn't a change "from ... to ...", but rather the introduction of a simpler capture-all (or capture-more-broadly) FD term, for the benefit of simplicity and contrast/complement with FC. AFD was not moved away from when FD was introduced.
Completely separately, just as something to think about in general (and I have no answers here at all) separately on the dixels / fixels thing: I often wonder whether bringing up "dixels" again at times ends up being helpful or rather keeps the difficulties of having to explain the notions and differences alive. But I don't know how that could be resolved though, because when the confusion arises, you want to explain it naturally. So it's a problem that keeps itself alive maybe. But maybe additionally because of that, there is potential value in keeping this entirely separate from AFD / FD.
Thijs
dchristiaens
Lestropie
jdtournier
This has come up a number of times, so could potentially do with a centralised reference to which to send people.
The change from "AFD" to "FD" happened kind of "silently", and the distinction between them I'm not even sure if there is an internal consensus. Here's my take:
"AFD" was originally defined based on the ODF amplitude in a specific direction; i.e. it's a dixel measurement. The "Apparent" makes sense just as it does for ADC.
"FD" came about in the finalisation of FBA. It's more explicitly defined as the ODF integral for a specific lobe; i.e. it's a fixel measurement. The "Apparent" got dropped as presentation of such was no longer principally about contrasting an SD-derived measure against the rest of the field, but instead about contrasting FD from FC and FDC within FBA.
Now some people use them interchangeably, and claim that the definition of "AFD" is wholly compatible with the fixel-wise quantity that is under use now as "FD". I think based on the original wording that's potentially erroneous. But I also think that attempting to drop "AFD" entirely due to the change in domain from dixel-wise to fixel-wise is maybe not wholly warranted. A dedicated documentation page would provide the space to elucidate the precise difference, where it changed over time in what publications, and the fact that linguistically we don't always provide a clean separation even ourselves.
I would also note that a visual distinction (#1128) between dixel-wise and fixel-wise quantities would be highly beneficial for the existing documentation page on fixels & dixels.