On Thursday, June 23, the Martinos Center’s Jon Polimeni will deliver a keynote lecture at the 2022 meeting of the Organization of Human Brain Mapping (OHBM) in Glasgow, Scotland. In advance of the lecture, Jon is an all-around fascinating guy, so we were thrilled when he agreed to submit to an “Ask Me Anything” (AMA) in advance of the keynote.
Read his responses to your questions below. And if you’re attending the meeting, don’t miss his talk tomorrow!
What should today’s brain mapper who uses fMRI know to get the most information from existing imaging technology? Do you have any tips and tricks for using the technology to gain the greatest neurobiological insight?
Great question! In my view, today there is enormous flexibility in how fMRI data are acquired in terms of spatial and temporal resolution, coverage, sensitivity, and image quality, mainly provided by technologies such as ultra-high magnetic fields, high-performance gradient coils, massively parallel receive coil arrays, and of course advances in pulse sequence and image reconstruction, and brain mappers should take full advantage of this. These technologies have all matured substantially in past years and are now ready for routine use. The additional capabilities enabled by these technologies provide many degrees of freedom in how data are acquired, and so while standardized protocols and techniques have clear value, I would encourage brain mappers to carefully consider what the specific needs are for any given study and then adapt their acquisition strategies and protocols accordingly.
A more concrete tip would be to try imaging at higher resolution. Higher imaging resolution can be helpful even for studies that are investigating large-scale brain organization because adequately sampling in space and time can help separate neural signal from physiological noise — e.g., by enabling clearer separation between gray matter tissue and surrounding cerebral spinal fluid, or by facilitating separation of task-driven activation from nuisance cardiac and respiratory fluctuations. Once the structured noise is removed, the data can be smoothed down to lower resolutions (spatially and temporally) to reduce thermal noise contributions. This can be counter-intuitive, but higher resolution acquisitions and appropriate noise removal can in the end provide better signal-to-noise ratios than conventional-resolution acquisitions.
But more broadly, I would encourage brain mappers to keep in mind the origins of the fMRI signals. All fMRI signals in use today are hemodynamic in origin, and are therefore shaped by the blood vessels of the brain. As I will discuss in my lecture, knowledge of the vascular anatomy and physiology can help interpret fMRI data, and help understand the subtle vascular influences on the data to gain deeper neurobiological insights from our experiments. What I would like to see more of in the future are auxiliary scans acquired during experimental sessions that seek to evaluate these vascular influences, either through measuring vascular anatomy through, e.g., the many forms of angiographic MRI that are available on all scanners, or through measuring vascular physiology through, e.g., simple calibrations using breath-hold challenges or even resting-state fMRI data.
What are the most significant technical hurdles limiting further improvements in spatial and temporal resolution in fMRI?
I’d say that there are two categories of hurdles that we are facing to further improve resolution: sensitivity limits and encoding limits. Because signal-to-noise ratio decreases rapidly with increasing resolution, higher resolutions are only useful in practice when sufficient levels of sensitivity are available, and often this is the main limiting factor. At the same time, it is not always possible to increase imaging resolution given the various constraints placed by our hardware and instrumentation, and often we are encoding limited in the sense that we are not able to encode the fMRI data at the desired spatial and temporal resolutions. Today, at 7 Tesla, we are able to achieve voxel sizes well below 1 mm in size, and sampling rates well below 1 s, thanks to the investment made into clever pulse sequence and image reconstruction design enabled by our advanced instrumentation and hardware, so we are currently mainly sensitivity limited I would say.
We could increase sensitivity by perhaps developing more sensitive coil detectors, but our array coils today are already very impressive. I am personally excited about the current renaissance of high-performance gradient coils, which provide more efficient image encoding that can help improve sensitivity. But moving to higher magnetic field strengths would be the most effective way to boost sensitivity, and there are many efforts across the world that are pushing >10 Tesla human MRI. Functional MRI at these extremely high field strengths will likely be encoding limited, and therefore will need innovative new approaches to increase spatial and temporal resolution to take full advantage of the sensitivity increase afforded by the increased field strength.
But overall, because the blood flow and oxygenation responses to neuronal activity appear to be regulated at impressively fine spatial and temporal scales in the brain, I think that today our imaging resolution is the main limiting factor, not the “biological resolution” of the hemodynamics, and this is likely to be the case well into the future.
You’ve been an ally since before there was a word for it. What would you advise your fellow male academics who aspire to be allies to their female colleagues and mentees?
All I would suggest to anyone would be to make an effort. I myself make an effort to check for my unconscious biases and then make corrections if needed, and in discussions that I lead I try to make space so that everyone can be heard. I would encourage others to try to do the same.
You’re also a musician and have played in bands. Do you have any favorite memories from your time with the bands?
Ha! Yes, I was in a band or two at some point in the distant past, and while I don’t have the exact spatiotemporal coordinates it was in my high school and college days mainly here in the Boston area. I played bass guitar (badly) and sang (even worse). We did have fun though.
I do have one favorite memory: at a first show with one band, at the end of our set we had an encore (thanks to our friends in the audience) but had run out of songs to play, except for one silly song that we played during practices as a warm-up, so we quickly decided to play this. I would usually start the song on bass and the rest of the band would then start playing with me, and I was so nervous that I started playing about twice as fast as usual, and the rest of the band had to follow suit. I had no idea how fast it was until afterwards when my bandmates where shocked by how fast we had played. I do think of this sometimes today when giving a talk, and due to some nervousness I start speaking faster than my usual pace, and only afterwards learn from the audience how fast it was. Some things don’t change. 😊