Neuroimaging analysis, like sex, is a fun, exciting activity sometimes resulting in death. Although it is an enjoyable pastime for responsible adults, however, each year sees the arrival of ever-increasing numbers of inexperienced youngsters who also want to participate. Often their interest is piqued by irresponsible professors and academic rakehells extolling the pleasures of cortex, neurons, and brain science; other times, their imagination is aroused by lurid images in seedy neuroimaging journals depicting impossibly large brain activations - often, it should be added, achieved without using proper correction, correction which most researchers piously encourage in public but secretly admit dampens spontaneity and feels unnatural.
The demand for instruction has consequently led to an industry of informal advice. Small wonder then the proliferation of blogs, websites, and toolboxes; some written in chummy tones, some begun in a white heat and then abandoned, some claiming to help the newcomer achieve a maximum of skill with a minimum of effort. (These latter ones, rightly recognized as fraudulent, say that the details should be left to the experts at Oxford and Bethesda; download this package, try out this sample dataset, press this button, and fluency will come as readily as a tree bears fruit. Maybe.)
But being able to judge where to begin can be difficult if not impossible - without knowing what one doesn’t know, how to choose? At any rate, what one does know is that there must be reliable guides out there somewhere; and if he is lucky enough to have a postdoc or an advisor teach him the fundamentals, he will at least have some idea of where to start. He will know in what general direction to look and avoid the easy, sleazy way of Googling vague generalities; or at the very least he will have mastered enough of the vocabulary to know how to make his questions more focused and precise. And if he is exceptionally lucky, he will be able to enroll in an introductory course.
How we all started learning fMRI analysis.
Such courses, unfortunately, are not always offered; or, what can sting more, they are offered, but not at the right time. Some times they are offered, but not taught well; or they are taught well, but only for those who already have a background in the subject. For his investment in the class our student, like a disappointed stockbroker, finds it hard to see anything but a deficit: he dutifully clicks buttons and recites definitions, but in private feels he hasn’t learned anything.
Discouraged, he begins to look for other options. One that comes to mind is attending a workshop. But even here, success is not guaranteed. To begin with, one applies by writing an essay about why they need to attend this particular workshop, keeping in mind the “soft requirements” listed on the website - which, turning out to be a misleading euphemism for actual requirements, demand that the student already be familiar with the topic he is trying to learn. This might seem unfair, but the organizers have their reasons: It may be an introductory workshop, but they need to assume a certain level of competence with programming, statistics, and imaging in order to get anything done at all. A course that assumes total ignorance would spend the entire week on the fundamentals, and practical exercises with imaging data would be so rare as to be virtually no help at all.
But what if he does have the background? What if he is exactly the type of applicant the organizers are looking for, and his application is accepted? Then there are fees and travel to think about. A brief look at the most popular workshops aside from AFNI (whose attendees do not pay for registration) reveals registration fees ranging from $200 to over $1,000. This in addition to lodging (sometimes provided for free, as at Neurohackademy) and travel, which, depending on the location of the workshop, can be prohibitively expensive. Travel grants can be applied for and money withdrawn from the lab’s funds to defray the costs, but in many cases the student will be paying part of the expense out of his own pocket.
In painting such a picture, my intention is not to discourage anyone from applying to these workshops. Far from it. There is no substitute for learning from the creators of the software that you will use during your career; and being able to talk with the developers in person is an experience that cannot be duplicated anywhere else. But the fact remains that those who are most in need of learning the basics have correspondingly low chances of being accepted to any workshop; and many of these inexperienced students, if they are lucky enough to attend, end up learning very little. And so they rejoin the already overflowing ranks of those who know just enough to make the software run, but, like the Sorcerer’s Apprentice, quickly find it beyond their control; and meanwhile the rest of academia remains puzzled why the rising tide of graduate students are unable to pull themselves up by their bootstraps.
Having been one of those students some time ago, I have written an e-book designed to help newcomers learn neuroimaging analysis: Andy’s Brain Book.
You may ask whether this has already been done before; and the answer is, Yes, it has. There are several other online resources available, from the introductory tutorials on the websites of AFNI, SPM, and FSL, to wikis compiled by laboratories all over the world. There are books, Githubs, blogs, and walkthroughs that one can find if he looks long enough. This particular book is not a new idea, nor is the approach radically different: the basic teaching method of exposition, development, and recapitulation, seasoned with examples and code here and there, is fundamentally the same.
What I do hope is useful, in any case, is the spirit in which the book is presented. The reader’s convenience is always kept in mind, as is the fact that students with different backgrounds are attempting to learn neuroimaging analysis. Some are psychologists, some are sociologists; some are adept statisticians, while others think a t-test is something done at the doctor’s office. Some are looking for the latest graph theory techniques, and others are just trying to make any graph of all - something that mom would put on the refrigerator, at least. Regardless of where you come from or what you know, Andy’s Brain Book is designed so that anyone from the rookie to the veteran can profit from it.
The book is organized into modules which are cross-referenced with each other as needed. An introductory module on Unix, for example, is designed to be used with AFNI and FSL, which both rely heavily on the Terminal; consequently it focuses on guiding the student toward an understanding of conditional statements and for-loops, which will be frequently used for analyzing several subjects at a time. When the reader then goes on to learn more about either AFNI or FSL, he will encounter reminders of other tutorials that he should already be familiar with in order to understand what will be coming up next; else, if he already knows the material, he can skip over the recommendation without any problems.
Example of cross-referencing in the book.
Screenshots are used to summarize blocks of text, and, where possible, the links are shown between what is entered into the graphical user interface and what is output to the Terminal:
Blocks of code are clearly marked with a light green background, and links to a Github page are provided in case the reader wishes to save the code as a shell script:
Where appropriate, videos are created to guide the student through complicated analyses that are easier to show on the screen than to explain in writing. Links to the videos are placed at the end of their corresponding chapters, and multiple videos related to the same topic are combined into playlists:
At the end of certain chapters, exercises are provided in order to consolidate what you’ve learned and to strengthen your neuroimaging muscles:
The platform for the book, ReadTheDocs, places the table of contents on the left sidebar, making it easy to stay oriented even when chapters are nested within other chapters:
Lastly, the search bar makes it easy to find topics based on even a single keyword:
In all cases, the reader’s needs are held in mind. The videos themselves have timestamps indicating when a specific topic is discussed, making it easier to find the information that is needed; the viewer may already know the background of the topic, but is looking for a certain command or illustration. The watchwords are Ease, Clarity, and Convenience. This is not to say that learning neuroimaging itself is easy, but that Difficulties are removed whenever possible. The path may be a hard one, but it should always be clear where one is going.
At this point I invite the reader to try it. Currently, the completed modules are AFNI, FSL, SPM, Unix, FreeSurfer, and E-Prime, and both diffusion and functional connectivity modules will be finished soon. Whether you need help with any of those packages, or whether you know someone who does - or, if you just want to see another perspective on something you already know - my hope is that the book will be a useful and convenient reference for those who need it. At the very least, it will serve as a guide for graduate students who are wrestling with strange new desires to analyze neuroimaging data, but who want to do it safely and with respect for their colleagues.