What’s the best name of our field and what’s the best attributing term for our data? There are many competing options: “Layer fMRI”, “mesoscopic fMRI”, “sub-millimeter fMRI”, “ultra-high resolution fMRI”, “laminar fMRI”, “cortical depth-dependent fMRI”. They differ with respect to how flashy they are, how scientifically appropriate they are, and how popular they are.
In this blog post, I want to review which ones are the most popular ones in the field and also share some thoughts on my favourite candidates.
Edits on March 3rd 2019 with contributions and clarifications taken from Kamil Uludağ, Sri Kashyap and Faruk Gulban.
Some terms describe the eventual goal of what the data will be used for, and some of them refer to technical ways they were acquired with. Some terms are intentionally ambiguous, too general, or even deceptively specific. However, mostly they mean the same thing. Namely, functional MRI data with unconventionally high spatial resolution to infer brain representations within the cortical GM sheet.
What do other people use
I tried to get a rough idea on the most popular terms in the field using two metrics. 1.) counting the number of publications and paper that use a specific terms, 2.) using google trends to find out which terms people type into Google the most, and Srirange doing a Twitter poll.
Requirements of a good term
- It should be well defined. The term that we use to describe our data should be simple and clear. It should not be too ambiguous or highfalutin. It should describe features of our data and idialy already imply what we want to use them for.
- It should be clear and catchy. The term that we are using to describe our data should not be too complicated. It would be a pity if it would be considered to be too technical by non-experts outside out research field. After all, our field of fMRI is in competition with other research fields. As such, if non-expert decision makers (in grant committees or conference symposium committees) do not fully understand how cool our field is, they might favor other fields with catchier names. For example, if the terms of other field promise cooler results, journal editors or conference program committee members might not give our field the attention it deserves (symposia or special issues). As such, the resting-state field promises information on “connectivity”, the DCM field promises information about “excitatory and inhibitory causal directional connections”, and the field of calibrated BOLD promises information about “quantitative metabolic demand”. So we shouldn’t be too modest about the information one could get from our data.
- It should be honest. The term that we use should not be too deceptive either. The term shouldn’t promise too much. Not at the beginning, at least. Otherwise the whole field might be under too much suspicion, which could hold us back.
- It should be agreeable. The best term is useless if it is not used by most of the people. A good term should help us find relevant abstract in search engines of conference proceedings and paper citation alerts.
Option 1: Layer fMRI
You might have guessed, this is my favorite term ;-). Its short, simple, catchy, and it implies what the data are aimed to use for.
In my understanding a layer is defined as a sheet of 3D-space over a surface. Thus, it’s a somewhat geometrical definition of describing a subdivision of space. So the term already nicely implies, how the data will be analyzed without being too specific which layer definition is assumed, or which kind of sample they refer to.
I have heard from some people that “layer fMRI” might be a too deceptive term, because it might suggest to the naive reader that we were sensitive to the cyto-architectonic cortical layers as used by Brodmann (which we are usually not). And thus, one should only be allowed to use the term “layer fMRI” when we have the resolution and specificity to see cytoarchitectonicaly-defined cortical layers. Personally, I found this gatekeeping argument rather weak. It would suggest that the term “layers” is reserved to the way how Brodmann used it. I don’t really agree with the view and think that “layers” is a rather general term that is used in many research and non-research related subjects.
The term “layers” is not protected for the use of Brodmann-defined cytoarchitectonic layers. Absolutely no sensible person would forbid a geologist to use the term “layers” to describe the structure of the earth. And similarly, there is no reason why one should forbid an fMRI researcher to use the term “layers” either. We should be allowed to divide the cortex in as many layers as we like.
Not even the term “cortical layer” can be claimed to refer to the six cytoarchitectonicaly-defined Brodman layers. There are 4 “vascular cortical layers” (a la Duvernoy), there are 5 neural layers (a la Cajal) and there are 5 “myelin cortical layers” (a la Theodor Kaas), so I do not see a reason why we should not be allowed to define our own “fMRI cortical layers” too.
While the term “cortical layer fMRI” is ambiguous enough that is remains unclear to which definition of cortical layers it refers to, it is specific enough to suggests that the resolution must be higher than the spatial scale of the cortical thickness.
One disadvantage of the term is that with the new trend of applying deep learning in fMRI, the deep learning layers might be misunderstood with the layers of cortical depth that we are looking at. So at conference proceeding search engines, I am getting more and more low resolution deep learning abstracts in my result list with the term “layer fMRI”.
Another disadvantage of the term “layer-fMRI” is that it is quite specific. The field of layer-fMRI is still very methods driven and most areas of interest for layer-fMRI researchers have a lot of overlap with columnar-fMRI researchers. So it would be nicer to have a term to describe both.
Whenever, I use that term “layer fMRI”, I always include a little disclaimer, E.g. in the footnote. Just so that I am not getting in trouble with the ‘terminology police’ that I am not referring to Brodmanns’ cyto-layers. It’s easy to be on the safe side here. Maybe one day this will not be necessary anymore.
Option 2: Laminar fMRI
In my early papers, I thought that I could avoid the whole terminology issue of “fMRI layers” vs. “cytoarchitectonical layers” by using the term “laminar fMRI” instead. Surprise surprise…. this didn’t work out. The term “lamine” is literally just the Latin translation for “thin layer” and thus can be used interchangeably as a synonym for “layer”. Also in the context of cytoarchitectonical defined “laminae”, this term is similarly often used compared to “layers”. Thus, I don’t see the term “laminar” a safer alternative to the term “layers” anymore. I see “laminar fMRI” as just being the more fancy Latin science talk alternative to “layer fMRI” that might be less understandable by a layperson.
The only small advantage of the term “laminae” compared to “layer” is that it’s easier to make an adjective out of it. “Laminar” (Latin for “of layers”) sounds much smoother than “layer-like” or “layer-dependent”.
Option 3: Cortical depth dependent fMRI
That term also didn’t last long. I think this is due the fact that it’s just too bulky. It’s such a stretch to use it in a comprehend-able sentence. Not even to mention how it looks when you want to use it in a catchy title.
Just for fun, say it out loud with me once: “Cor-ti-cal depth de-pen-dent f-M-R-I”. This is such a mouth-full. Imagine a twitter handle like this. The character count would be exceeded before you even start wringing it 😉 Its so long that it doesn’t properly fit in a Venn diagram (see first Fig. above).
I think the bulkiness of this term is also the reason, why there is virtually not a single google query with this term (Fig. above), even though every 15th publication contains this term.
I am also not really sure about the word “depth” in this term. Maybe it’s only me, but it somewhat sounds like a distance from a surface. It doesn’t explicitly say anything about how the “depth” is estimated but to me, it implicitly suggests that it is estimated based on an equi-distance depth sampling approach. I found it much easier to imagine a “layer” to have spatially heterogeneous thicknesses than a “depth”.
Hower, as nicely pointed out by Kamil Uludağ: “Only the latter [cortical depth dependent] has a nice abbreviation (even though we always have to write “signal” or response” after BOLD, because it is an adjective). In addition, cortical depth does not necessary signify equi-distance as it can be calculated accounting for curvature.”
Option 4: Mesoscopic fMRI
The term mesoscopic or meso-scale has gained quite some popularity in the field in the last years. It is more commonly used by big players in the field including Rainger Goebel, David Feinberg, Serge Dumoulin, Nick Weiskopf, Natalia Petridou and Jozien Goense. The term “meso” literally means intermediate.
In this paper, it is concluded that the boundaries of are somewhere between 0.1 mm and 0.5 mm, however noting that there is no consensus yet.
I love and hate the term for its ambiguity. Since it’s usually not explicitly stated to which scales it is intermediate to, it can mean anything. Thus, I like it because it nicely combines layer- and column-fMRI. However, it is so ambiguous that people from other fields might be confused. I am a trained physicist, and I always cringe a bit when I hear the term “mesoscopic fMRI”. In mesoscopic physics, the term “mesoscale” refers to object of the size between nm and μm, whereas in mesoscale meteorology it refers to features in the range of 5km and few hundred km.
While I do not see any clear reason, why we should not be allowed to define the word “meso” for the context of fMRI too, I find it not specific enough still. As such, even in the realm of fMRI it could be misunderstood to refer to the mesoscale between 3mm (cortical thickness) and 3cm (cortical areas), such as is retinotopic or somatotopic fMRI.
In other disciplines (outside MRI), the term “meso” usually refers to a scale that lies somewhere in the middle, between the smallest possible units and the largest possible units. In the realm of fMRI, however, this is not the case. Here, “meso-scale” refers to a spatial scale that is clearly on one extreme side of the spectrum; the smallest voxels currently possible.
The term of “mesoscopic fMRI” has the advantage that it refers to the neuroscience investigation as nicely pointed out by Kamil Uludağ.
Option 5: Sub-millimeter fMRI
The term “sub-millimeter” might be the best defined one among all of them. There is no question, which data-set qualifies to be classified to have this term and which data set doesn’t qualify. And this is simply based on a clear physical unit, mm that can cannot be easily redefined for the context of fMRI.
Even though, I used this term a lot, I find it both too specific and not specific enough.
- It’s not specific enough because it does not say anything about the goal of the research direction.
- And it’s too specific because it might exclude a many special cases of layer-dependent fMRI data that didn’t happen to be acquired with sub-millimeter fMRI. E.g. in lissencephalic animal brains (without gyrification), it is common to use slice-thicknesses that are not in the sub-millimeter regime. The in-plane resolution however, can be quite exquisite and allows much clearer layer-fMRI analysis than “sub-millimeter fMRI” in humans. Another example would be the very thick human motor cortex (4-5 mm thickness), where super-millimeter voxels can provide enough resolution for a decent layer-fMRI analysis. Thus, the term ‘sub-millimeter’ feels a bit too exclusive to me.
Option 6: high resolution fMRI
The term “high resolution” might also be a good alternative (Thanks to Sri Kashyap for pointing this out to me). I feel that this term is not really overstating anything and it thus one of the most honest terms. The term “high resolution” would refer to any resolutions that is higher than the conventional 2-3 mm of most fMRI studies.
With this in mind, it might not really be specific enough for the datasets that aim to investigate intra-cortical signal distributions. Anything with higher resolutions than 2mm might also be classifiable as “high-resolution fMRI”. In fact some of the top google scholar hits of papers that have “high resolution fMRI in the title) refers to resolutions of 1.5 mm – 2.0 mm (Example 1, Example 2, Example 3).
Option 7: Ultra-high resolution fMRI
I believe the term of “ultra-high” originally came into the realm of MRI and fMRI from its engineering side. Since, MRI is using radio waves, MRI researchers also adopted part of the corresponding RF-engineering terminology. In the designation of radio frequencies, the term “ultra high frequencies” refers to the frequency band between 300 MHz and 3GHz. And coincidently, the lower end of the spectrum refers is to the field strength of 7T. And consequently, in MRI the field strengths of ≥ 7 Tesla are often referred to as “ultra high fields”.
While this definition is pretty official and even enforced by the corresponding United Nation agency ITU, the exact borders of the respective frequency bands are arbitrary and lack a physical justification. It is just based on historically agreed standards. Thus, I believe there is no reason, why our field should not be allowed to do the same for fMRI and make our own arbitrary definitions of what “ultra high resolution” refers to.
Since the term “ultra” already feels pretty superlative, the usage of such a term automatically begs the question of what comes above it (e.g. if one day in the far future, when cellular fMRI will be possible). Looking at the terms in the RF spectrum (Fig. above) also suggests that there is a lot of room towards even more extreme resolutions. The next stages of even higher resolutions could be called “super high resolution”. And as soon as that term is worn out, we could use the even more extreme term of “extra high resolution”.
The future development
But what can come after “super high” and “extra high” resolution? If I orient myself on the terminology in the Sci-Fi genre, we don’t seem to be running out of even more extreme superlative words. Maybe “hyper resolution” or “mega resolution” will be the terms of the next generation scanners?
One trend that has actually started already is to use the existing terms with some extra accompanying addition (e.g. from the Pittsburgh and Maastricht groups). E.g. true laminar fMRI or real layer fMRI. I am not sure if it might be a bit exclusive, however. Does it suggest that everybody without that little addition it doing un-true layer-fMRI?
Of course, the term of “true laminar resolutions” is fully defendable, as it refers to acquisition and not the results. If, for the sake of argument, a layer is 200 micro, how would you differentiate an acquisition with 800 micro and 100 micro (as in this study by Kaskyap et al.)?
Comments from other people
So which term should we use? Based on the discussions above, I think that all of the terms are usable and defendable. As nice as it would be, to have a clear brand-name that the entire field is agreeing on, I do not find a good reason to enforce specific terms over others and perform this kind of gatekeeping.
The term “laminar fMRI” seems to be quite popular already, so it would make sense to settle on this one. However, since I committed already to “layerfMRI” with my twitter handle and my blog URL, it would result in some inconvenience and inconsistency from my side
I want to thank multiple anonymous reviewers for making me think about these terminologies. I thank Kamil Uludağ and Sri Kashyap for very detailed comments on the blog post that lead to corresponding revisions. I thank Faruk Gulban for point me to the definitions of Mesoscale. I want to that Jonathan Polimeni for our discussion on the topic at the BIDS meeting in Magdeburg 2018. I want to thank Valentin Kemper for his lecture on the history of “ultra-high” frequency. I also want to thank Faruk Gulban, Ingo Marquard, and Marian Schneider for corresponding discussions over lunch and coffee brakes.