The History of Brain Imaging Data Structure (BIDS)

One development in neuroimaging research that everyone should be aware of is the advent of Brain Imaging Data Structure (BIDS). This was intended to correct the old ad hoc way of organizing data, which back in the day meant doing whatever the hell you wanted. No organizing scheme was too hare-brained; no folder labels were too outlandish. Here is a sample of what such a scheme looked like during those benighted times; read, and despair:

• BrainData

• OnsetTimesRevised

• MyFirstAnal

• GroupAnal

These names, which I definitely did not use during my post-graduate years, are not only ambiguous, but can mislead. For example, MyFirstAnal (which, for the record, is shorthand for “Analysis” (I hope)) clearly contains some kind of analysis; but which one, and why is it the first? How many did the researcher go through before hitting upon the one that worked, and which ones were used for the GroupAnal(ysis)?

With such harum-scarum labels, slip-shod and slap-dash, as well as many other denigrating hyphenated words whose very structure seems destined to denote confusion and general sloppiness, it is no wonder that back then so many people were stressed out, perplexed, and angry; unlike today, in which everything is neatly organized, clean, and well-run, and researchers are consequently pleasant and amiable. It is a wonder our forebears were able to do anything at all; and it may be safely assumed that everything written, thought about, or said up until about five years ago is rubbish. Under the new dispensation, we have finally found truth.

What brought about this change? About a decade ago during a meeting at the Organization for Human Brain Mapping conference, a group of researchers met to discuss whether data organization could be standardized; the welter of idiosyncratic names would be unknotted, the babel of languages would be reduced to one, and everybody could communicate more clearly. Reading another’s code would be much easier, and applying someone else’s script would require only minor edits to the file names; the rest of the folders would more or less be the same. No longer would the message board moderators at AFNI suffer apoplectic shock when trying to decipher a block of code rendered unintelligible by a naming scheme both deeply personal to the creator and utterly baffling to the outsider. Everything would be streamlined, simplified, and easy to understand.

Years later, their efforts have been fruitful, with BIDS widely adopted by a variety of data repositories and software analysis packages. For example, any data organized in BIDS format can be read and imported by the CONN toolbox, and this process takes considerably less time than doing it manually, as well as reducing the potential for human error; a website like OpenNeuro.org requires any uploads to be in BIDS format, and instructors who use it for courses or workshops can rest assured that their scripts should work for nearly anyone; preprocessing pipelines such as fMRIPREP require BIDS format, and it is likely that any new tool or software package intended for a wide audience will have to adopt it as well.

Although the use of BIDS may seem irresistible, there will always be some disaffected researchers who kick against it. Who came up with this convention, they say, and why should everybody have to use it? Virtually all labs have had their own way of doing things since their inception, and some may see no reason to change. Moreover, some people may simply not like the way that BIDS is laid out, and they may have legitimate grievances about the naming conventions. One size does not fit all, they say, and by having so many websites and software packages not only encourage but require BIDS, this in effect forces a kind of monopoly on the marketplace of neuroimaging research - one that they neither wanted nor asked for.

These complaints are valid, but at the end of the day, BIDS was not created to cause difficulties or confusion for anyone; it was a response to a long-standing, general level of discontent with how data was organized. At last a group of people decided to address it, and the rapid and widespread adoption of their standard indicates that many people agree with the severity of the difficulty and the usefulness of the proposed solution. Whether you agree with it or not, it looks like it’s here to stay, and the modern researcher needs to know how to convert their own data into BIDS, and how to use BIDS data from other groups. Such is the evolution of any professional field, and the modern researcher must adapt.

---

More details about the history of BIDS can be found in the preprint, which has been accepted for publication and should be available soon.

Neurodesk: The Deluxe Neuroimaging Suite

“I sometimes think that writing is like driving a sheep down the road. If there’s any gate open to the left or the right the reader will most certainly go into it.”

— C. S. Lewis

Teaching a neuroimaging workshop is similar. If there is any boredom, any lull in the action, you are already at risk of losing your audience; but the one thing guaranteed to sink their attention is a student having problems with their software while you are giving a follow-along demonstration. Up goes a hand; then comes a voice, perplexed and slightly annoyed, while they squint at their screen with the same unprepossessing facial expression they would make while passing a log of feces the size of a foam roller: “It says there’s a problem with…” followed by any number of potential errors:

• There is a path conflict between that software package and another one installed on their machine;

• The student made a typo in the Unix terminal that caused a chain reaction wiping everything off their system, including their trusted homemade recipe for Metamucil;

• The computer is not, technically, on

At this point the demonstration grinds to a halt, and the audience is jarred by the stopped momentum, their minds feeling like riders on a bus that has suddenly hit the brakes to avoid hitting a man sprinting across the street trying to get to the closest restroom before having an unforgivable accident. Once the passengers understand that it may take a while to get moving again, their attention wanders off through any of the numerous gates nearby: phones, neighbors, bathroom break (turns out their facial expression meant more than one thing); and - take it from me - it is hell not only trying to gather together their attention again for another collective push, but to reset their minds into the same train of thought that you tried so hard to establish in the first place.

Why does this happen? Consider the two primary workshop setups: On the one hand, we can have everyone bring their own computer, and install all of the software on it. On the other hand, we can conduct the workshop in a computer lab, in which each computer is identical.

There are advantages and disadvantages to each. If everyone brings their own computer, they understand how to analyze the data on their machine, which they will probably be using for most of their analyses; but at the same time they run the risk of path issues, or conflicts with other software, or they simply may not have the right computer. (I’m sure you have encountered those sad souls who try installing AFNI on a Windows machine - no matter how many patches and other tricks they try, it simply doesn’t work that well.)

If the workshop is held in a computer lab instead, you get around the problems of path conflicts and any other issues with the software, since everyone is using the same machine. But then again, they probably will never use those computers once they leave the room, and when they return to the computers they usually work with, they will probably be at a loss as to how to apply what they learned to another environment. Sad!

These problems have been known for some time, but no matter how much we complained about them, nothing ever got better. That is, until the release of Neurodesk.org, a web-based neuroimaging suite. This is a website that can be accessed from anywhere, whether you are using Windows or Ubuntu, and the environment will be the same for everybody. Imagine a computer that you can access from anywhere which contains all of the neuroimaging software you could ever need, from different versions of AFNI to MRtrix to FreeSurfer, and beyond. It may not be neuroimaging nerd paradise, but it’s pretty close.

The idea is the same as a container - and Neurodesk can in fact be downloaded as one. Containers such as Docker have been around for a while now, and one of its most popular uses in the neuroimaging community is to enable fMRIPREP, a preprocessing pipeline that draws upon libraries and commands from many different software packages; putting them all into a single container segregates them from the rest of your computer, and prevents any issues with paths or different software versions. The advantage of Neurodesk is to not only provide an extensive suite of packages covering every type of neuroimaging analysis, but to allow you to use them within a web browser:

Example screenshot of the Neurodesk environment, launched from a Google Chrome browser. The Functional Imaging section, for example, lists a dozen software packages, from the more popular AFNI, SPM, and FSL, to lesser known packages such as laynii and niistat. The most recent half dozen versions of the package are shown in the FSL menu. Loading any of these packages is quick, easy, and reliable.

I have just begun working with Neurodesk, but I can already say that it passed the first test with flying colors: I was able to lead a small group of students through a diffusion imaging analysis from start to finish, and nobody had any accidents. They also didn’t have any problems with the software.

There will be a more comprehensive report to come, once I get more experience with Neurodesk and can come up with a demonstration useful for newcomers. In the meantime, eat your fiber, stock up on plenty of Charmin, and let me know what you think about Neurodesk in the comments below.

Meta-Analysis of fMRI Data

Does it seem to you that there is a new fMRI study every year? If so, you’re wrong - there are in fact hundreds, perhaps thousands. And the good men and women carrying out this research know that each study, by itself, is a small piece of the puzzle; it isn’t likely that any one of them exactly located where, for example, the brain lit up in surprise when O.J. was acquitted. (It was the limbic system.) Add to this the phenomenon of statistical power, in which certain studies may report an effect by virtue of having enough subjects to detect the effect, while another study with relatively fewer subjects is unable to find it.

There are other factors as well - smoothing, normalizing, and any other interpolations of the brain data will warp, distend, and reshape the brain. And although those operations are necessary to increase power and to place all of the brains into a standardized space, at the same time it creates more uncertainty about where activations are located.

Luckily, with the remarkable increase of neuroimaging studies over the past two decades, and the decision by most researchers to normalize their data to the common space of the Montreal Neurological Institute brain templates, we can leverage the number of studies that have been published to visualize where there is significant overlap between the results. If many studies report the same voxels as being significant or having strong effects, then we can be more certain that those voxels really are active. Instead of measuring the variability of individual subjects, we are measuring the variability of individual studies; the total number of studies in our meta-analysis is a type of group analysis.

The idea behind meta-analysis is simple, although, like most ideas that appear simple to us now, it took considerable insight to create. In one of the earliest instances of meta-analysis, English statistician Karl Pearson combined the results from several different studies examining the effect of inoculation on enteric fever, also known as typhoid fever, in different parts of the British Empire. By calculating the correlation between inoculation and mortality (in this scenario, denoting survival rate), Pearson concluded that, on average, there was a noticeable positive correlation across the testing sites; and that this average was probably more informative than looking at any individual site.

Table 1 from the original Pearson 1904 paper, comparing across studies the effect of inoculation on enteric fever.

To understand why, we should also look at Pearson’s calculation of “probable error” within each site as well - a term which we would today probably call “standard deviation”, or a measurement of the variability within each sample. Not only do the averages within each sample differ from each other, but so do their variabilities; and the Ladysmith Garrison sample even has a negative correlation between mortality and inoculation, against the trend of all of the other samples. In other words, if someone just looked at the Ladysmith Garrison sample and concluded that there was a negligible or negative correlation, he would be misled about the average trend from all of the other samples.

To explain these deviations (or “extreme irregularity”, as Pearson would say), Pearson judged that other factors were possibly at play, such as the environment in which the inoculations were administered - which could mean anything from the local climate’s effects on the durability of the vaccine before the advent of reliable refrigeration, to the competence of the doctors administering the vaccine - or the self-selection bias of the volunteers, which could be a confound. In any case, if we consider that many of these confounds would cancel each other out across many studies, we can assume that the average across the studies is a more reliable indicator of the effect. (For more details on the history of meta-analysis, see this paper by Kevin O’Rourke, and for a more in-depth statistical overview of the Pearson paper, see this article by Harry Shannon.)

Over a century later, we apply these same principles to fMRI studies. Now that we have access to hundreds or possibly tends of thousands of datasets for a given effect, we can average across the data in each voxel in standardized space to generate meta-analysis maps - and to do that, we can choose from several popular software packages and websites, including GingerALE and Neurosynth.org. GingerALE, for example, was designed to be used with brainmap.org, a neuroimaging database that curates datasets and allows the user to download a subset of them which focus on a given cognitive process. Neurosynth, on the other hand, analyzes thousands of imaging maps on the fly, requiring just a few keywords that correspond to whichever effect you are interested in. And for those looking for a tool that can search for rare or even non-existent search terms and phrases in the neuroimaging literature, Neuroquery uses a sophisticated semantic smoothing algorithm to add weight to semantically related terms.

Example screenshot from Neuroquery.org, using the search term “self referential processing”, which is not commonly used in the literature. Neuroquery uses semantic smoothing to predict where the activation for this search term will probably be located.

You can learn more about how to use each of these by visiting their respective websites; I have also started to write some tutorials about how to use each of them, which you can find here.

Trends in Best Practices for fMRI Research

If you are a newcomer to the field of neuroimaging, you may find bewildering the range of software packages, methods, and concepts in the field; aside from learning some of the basics of fMRI analysis, perhaps, or how to analyze an EEG dataset from start to finish, you may have questions such as:

• What are other, more experienced researchers doing?

• What is the best way to organize and analyze my data? Is this BIDS thing for real, or just a fad?

• Will univariate analyses be around for a while, or will they eventually be replaced by multivariate techniques?

• Was O.J. guilty?

It is natural to ponder all of these, and more, as you advance along your career as a neuroimaging researcher. Although it’s impossible for anyone to answer all of these with complete certainty, we can make some educated guesses about the direction of the field as a whole, including how results are displayed and reported, what statistical techniques are considered necessary, and what other tools the modern researcher should have in their toolkit. All of this, and more, I discussed in a talk hosted by the University of Connecticut, which you can watch below.

University of South Carolina Hackathon

Last November I was invited to a “Brain Imaging Hackathon”; a term not entirely clear, but carrying connotations of programming, hacking, and staying up all night discussing neuroimaging, the brain, and the latest developments in the field. It may also bring to mind a conclave of skinny, pale, socially awkward nerds who have nothing better to do with their free time.

I found out that it in fact means a conference attended by many attractive, confident, athletic people who happen to be researchers and computer programmers discussing their latest projects - in this case, focusing on web-based tools for visualizing and analyzing neuroimaging data. For example, NiiVue is a web-based imaging visualization tool that allows neuroimagers to share and examine brain volumes without having to upload the volumes themselves to an online database, or send them over email; Boostlet.js contains several “boostlets”, or bookmark apps, which can be used to segment a brain image from most web databases or plot the histogram intensity of the image, with just a click. And EZbids, part of the brainlife.io website, aims to streamline and simplify the conversion of neuroimaging data into BIDS format, a relatively new data organization standard that is becoming rapidly adopted.

To my delight, I was also introduced to a new website called Neurodesk, an online library of neuroimaging software packages and other resources that a researcher can use to analyze his data in a container without having to install or download anything - virtually every analysis package, from FSL to MRtrix, is available to use, and works almost as quickly as it would on your local machine. The advantage is that you no longer have to worry about conflicts in the paths between different packages, or be concerned about whether the latest version of Python is compatible with the software you are trying to install; we have all felt the sudden rush of panic when the Terminal informs us that Python version 3.1.2 is no longer supported, and that absolutely nothing you do can fix it, ever - similar to the feeling you get when you accidentally flush your car keys down the toilet at a truck stop in Walla Walla, Washington State, or when you are so hopped up on sugar at the movie theater that you use Google Maps on your phone to try to locate the restroom.

All of this, along with interviews with the developers of these packages, I was able to record during my trip, which you can review in the video below. I plan to record more of these as I travel around the country for workshops and conferences, talking with some of the leading developers in the field and discovering what the latest trends are in neuroimaging analysis - so make sure to stay tuned, subscribe, and if you find my keys, let me know.

How does Magnetic Resonance Imaging Work?

For anyone who has tried explaining MRI physics to the layman, the expression on his face follows a very particular progression: First the eyes are narrowed and attentive, the brow slightly furrowed, as you speak of water and hydrogen, blood and oxygen, tissue and bone. These are tangible, they are real; the man can feel them on his own body, or he has an easy enough time picturing them. His look of concentration wavers a bit once you talk about spin, and how it’s an intrinsic property of all atoms, how it’s both like and unlike the spin he experienced as a child on the merry-go-round. And he can be forgiven for looking puzzled when you describe spins as either aligning with a magnetic field or aligning in the opposite direction, partly due to the field but also partly due to chance, and that these spins, in and of themselves, are either up or down; they do not pass through some intermediate stage. And that although we have many figures and paintings of spinning electrons, we evidently draw from memory as we do of a distant loved one; as the electron is a very shy lady indeed, and no one has ever taken her picture or seen her in the flesh.

But that shadow of doubt is gone in an instant, his demeanor ready for more, once you begin talking about magnets. Magnets! Everyone has played with them; everyone understands intuitively the nature of the poles, attracting their opposites, repelling their identical twins. Everyone has observed them acting through solid matter: Tables, books, hands; none of these stop the magnet from pulling on filaments, metal, other magnets. An invisible force, whose effects are plain as day. It is only when you begin talking about gyromagnetic ratios and resonance that his mind begins to falter. Yes, the atoms spin at an incredible rate; yes, we can push them periodically just as we would a child in a swing, tilting the atoms on their side. And then this potential energy is released, and the signal is picked up by sensitive recording devices inside the scanner. So far, so good.

But magnetic gradients? K-space? At this point our listener’s inner eye becomes clouded over. There is something about Fourier transforms, and how each point in k-space corresponds to the magnitude of the image - or was that the contrast? In any case, he will attempt to understand it the next day, or the day after that; but it invariably comes to pass that our thinker finds himself frustrated, and, not seeing any purpose to continue wasting his time trying to understand - unless he is a very eager student indeed - he quits.

Nobody would claim that a video would clear up all of his confusion, but it might go a long way toward making MRI physics more accessible. The video above contains an impressive illustration of how MRI machines work, a brief but effective description of MRI physics, and an animation of how images are reconstructed from k-space. I recommend this to any student who has found himself bewildered by the topic, and I hope that it helps everyone appreciate just how complex and wonderful these machines are.

Now if we could just get a picture of that electron.

MRtrix Fixel-Based Analysis

One of the more advanced features of MRtrix is Fixel-Based Analysis (FBA), a technique to measure both the fiber density and the fiber cross-section of a given piece of white matter. The developers of the package invented the term “fixel” to rhyme with “voxel” (kind of), indicating that they both contain values representing a metric of brain activity or brain structure. The typical voxels we think of contain a single number representing contrast - either the contrast between grey and white matter or other tissue types, in the case of a T1-weighted anatomical image, or the contrast between the intensity of the BOLD signal in a T2-weighted functional image.

Fixels, on the other hand, are MRtrix-calculated values that are stored in voxels; they are the smallest unit of resolution for measuring white-matter related metrics, such as fiber density or fiber cross-section. These terms are defined in more detail in the Raffelt et al. 2017 paper, in which fiber density refers to the overall number of fibers compressed into a single voxel. Fiber cross-section, on the other hand, refers to the amount of the voxel that is occupied by the fiber bundle. These differences are illustrated in the following figure, taken from the Raffelt et al. 2017 study:

The goal of Fixel-Based Analysis is to compare groups and determine which fixels show a difference in fiber density, cross-section, or a combination of the two (referred to as Fiber Density & Cross-Section, or FDC). Many patient populations, such as persons with Alzheimer’s or other age-related dementias, have markedly different fiber density and cross-sections in major white-matter pathways, and the technique described above is a way to visualize and quantify these differences.

I have written a tutorial demonstrating how to do this for the BTC Preop dataset, available on OpenNeuro, which includes glioma patients as well as controls. As there are 36 participants in total, I recommend running this analysis on a supercomputing cluster. In fact, you probably won’t be able to run this analysis without a computing cluster, because with 36 datasets, commands such as “population_template” and “fixelcfestats” can take dozens if not hundreds of hours to run. The datasets that are generated are also huge. All of this points to using a powerful supercomputing cluster with plenty of storage in order to run the analyses, and then downloading the final product to visualize on your local computer, or mounting a volume of the computing cluster on your machine.

The supercomputing code for Fixel-Based Analysis, adapted from the code outlined on the MRtrix FBA tutorial page, can be found here. The tutorial may be updated to reflect better supercomputing practices - for example, using an array instead of creating an individual template file and submitting it for each subject - but it should work for most purposes.

CAT12 Tutorials

Last summer, I uploaded a few tutorials about how to use CAT12, a Matlab toolbox for estimating cortical volume. Due to a data use agreement, I have had to remove any identifying subject information, such as scan ID, scan date, age, and sex; after these edits, the original videos are back online, in slightly edited form.

The current playlist can be found here, and the e-book chapters can be found here; the next video in production focuses on how to do this type of analysis on the Great Lakes computing cluster at Michigan. Stay tuned for more updates!

University of Washington Workshop Complete and Other News

Last month we had a workshop at the University of Washington in Seattle; we focused on functional connectivity with the CONN toolbox, and different types of machine learning with The Decoding Toolbox, along with how to avoid pitfalls with using these packages. You can find the details here.

Looking forward, once I finish teaching for the semester here at the University of Michigan, in May I will be traveling to the University of Florida to give a workshop on AFNI, along with a special session about how to analyze animal brains in particular, and then shortly after that I will be hosting a workshop at the University of Wisconsin-Milwaukee on machine learning. In the latter workshop, I hope to have a working demonstration of hyperalignment, a new iteration of multi-voxel pattern analysis out of Jim Haxby’s lab, which transforms voxel-level data to a higher-dimension abstract information space, which can markedly improve classification accuracy. These are still under construction, but check the Workshops page for updates. And if you would like to have me host a workshop at your institution, and, more importantly, become part of the hall of fame, please feel free to get in touch.

Other Updates

In other news, I am updating the MRtrix walkthrough on the e-book to include Fixel-Based Analysis, which generates whole-brain maps of fixel-based values such as Fiber Density and Fiber Cross-Sections; this chapter, which is still in progress, can be found here. These maps are similar to TBSS’s diffusion tensor metrics, but unlike tensors, they are able to deal with crossing fibers. And, since I believe that MRtrix is becoming more widespread and will eventually surpass TBSS in popularity, graduate students should learn how to use it in order to stay current with the most recent methods.

I am also working with one of our tech specialists here at the University of Michigan, Bennet Fauber, to create a walkthrough about how to use the Great Lakes supercomputing cluster here at Michigan. This is an especially important skill for graduate students and all researchers to learn; if you are analyzing larger datasets, you may need to use a computing cluster to finish the analyses in a reasonable amount of time, and to store some of the intermediate results. Bennet’s tutorial sketches can be found here, and they will continue to be updated in the coming months.

While the e-book is still being regularly updated, I will include more informal updates here on the blog as well - what is in the works, where I am traveling, Novak Djokovic’s progress, and so on. More to come soon.

Updated: FreeSurfer Job Submission on Open Science Grid

A few years ago, I created a tutorial about how to use a supercomputing cluster called the Open Science Grid. The supercomputer was available to all researchers affiliated with a United States institution, and it was easy to analyze large numbers of FreeSurfer subjects in a short period of time; for those whose analyses were restricted mostly to FreeSurfer, it was like having your own personal supercomputer. Once the data was analyzed you could then download it to your local machine and use it however you like.

In 2019 someone told me that FreeSurfer was no longer being supported on the Open Science Grid; consequently, I removed the video I had on YouTube explaining how to use it, and put a disclaimer on the corresponding chapter of Andy’s Brain Book. Recently, however, alert reader Christina Koch let me know that although fsurf has been deprecated, there is a new way to submit FreeSurfer jobs to the Open Science Grid. Here is what she wrote on my e-book:

“fsurf has been deprecated and is no longer supported; to run Freesurfer jobs on the OSG, you can request an account on the OSG Connnect service (available to all researchers affiliated with a US-based institution or project) and then follow the instructions on this page which shows how to submit your own Freesurfer workload to the OSG.”

I haven’t gotten the chance to work with it yet and make an updated video about how to use it, but I encourage everyone to check it out. If there are any issues, let me know before I begin trying it on my own.