Ivan Kirov, PhD, is an associate professor of radiology and neurology at NYU Grossman School of Medicine and a scientist at the Center for Advanced Imaging Innovation and Research at NYU Langone Health. Dr. Kirov’s work is primarily focused on the development and application of magnetic resonance spectroscopy techniques to the study of neurological disorders, such as traumatic brain injury, Alzheimer’s disease, and multiple sclerosis. A graduate of NYU School of Medicine, he now teaches in NYU Grossman’s biomedical imaging and technology PhD program. In 2023, Dr. Kirov and Assaf Tal, PhD, received a five-year $3.8 million award from the National Institute on Aging to investigate cellular viscosity as a marker for Alzheimer’s disease by using positron emission tomography and advanced magnetic resonance spectroscopy. Our conversation was edited for clarity and length.
How does magnetic resonance spectroscopy (MRS) differ from the kind of imaging that most people associate with MRI?
Spectroscopy measures levels of metabolites and it looks at metabolites instead of looking at water. Everything else that we do with MRI uses the water signal: structural MRI is what water inside the body looks like; diffusion MRI is about how the water behaves. But with spectroscopy you look at chemicals.
And from spectroscopy you get information that isn’t usually available in an MR image.
Information that’s never available in an image—it’s entirely unique. A spectroscopic exam tells you about levels of metabolites independent of the anatomy. If you’re doing MRS in the brain, you’re looking at brain chemistry as opposed to macroscopic structure.
There’s one metabolite to which spectroscopy owes a lot of its usefulness called N-acetylaspartate, or NAA for short. It’s a metabolite whose signal is thought to originate almost entirely from neurons. In general, the higher the NAA levels, the better the neurons are doing. In most neurological disorders, the levels of NAA are lower, indicating some sort of neuronal disfunction. Notice that we’re talking about a specific cell type, which is very unique for an imaging method.
Over the course of your career, you have used MRS to study neurological disorders, with particular focus on traumatic brain injury, Alzheimer’s disease, and multiple sclerosis. What led you to these neurological disorders in particular?
I started working on traumatic brain injury, or TBI, in graduate school, and when I’m interested in a problem, I like to pick it apart methodically, so I’ve continued in this area even as I expanded into other topics along the way.
There’s a lot of public interest around traumatic brain injury.
Most people have at some point hit their head and had a TBI—yes, most people. Keep in mind that you don’t have to lose consciousness—you can be dazed or seeing stars, and that’s a concussion already.
What can MRS tell us about a condition like concussion?
By giving you levels of chemicals, it can tell you about the health of particular cell types in the brain. And the idea is that by knowing what type of brain chemistry is normal, you can understand what’s wrong: whether neurons are involved, whether astrocytes are involved, or if there’s a problem with energy metabolism or membrane turnover—these are the four main markers that MRS can measure. All this information, coupled with other techniques, can give clues as to where the damage is hitting.
Is that kind of information currently used by doctors?
It’s not part of routine protocols—it has to be ordered by the referring physician for a specific purpose. For example, you see a lesion on a T1 or T2 weighted image, but most lesions look the same and you have doubt as to what the condition could be. So, you can order MRS to help in the differential diagnosis: is it disease A or disease B? The chemical profile can help.
In your research on MRS in traumatic brain injury, are you more focused on answering fundamental questions about this condition or on trying to provide tools to clinicians?
There always has to be a clinical application in mind, if not on the level of an individual patient then on a group level that is useful for patients. For instance, in clinical trials if you want to take a certain patient population with a certain risk factor, you may be able to use magnetic resonance spectroscopy to elucidate that. So far, most of the research has contributed to understanding the disease on the group level; much less on the level where markers would be widely used. That’s a long-winded way to say that a goal of my research is to find exactly that: radiological biomarkers that can be helpful to clinicians.
What are the major challenges in getting MRS techniques to the clinic?
The challenges are of course specific to the diseases you’re working with, but one major challenge common to all techniques is that in clinical research with humans there are too many variables constantly changing from experiment to experiment.
So, it’s about reproducibility, standardization of protocols, and the many small details that make studies comparable in large sets.
Exactly. But it’s also more than that. Every patient is different—a person is not like a mouse model that can be closely replicated. For example, in TBI, everyone hits their head in a different way, and some people don’t have any symptoms while other people are very impaired. And there are pre-existing conditions, psychological factors, age, and we don’t even know whether the same thing happens after different kinds of impacts. Even where and how you look in the brain to see this injury is very important, because there are many different ways to use the same MRS technique. Plus, if you don’t know what the nature of the injury is, you have to experiment with different methods of postprocessing the data to explore where the injury may be, and everybody does this differently…
In an NIH-supported project you co-led with Guillaume Madelin, you used sodium MRI and spectroscopy to study long-term outcomes in TBI. Does the high heterogeneity that you just described mean that when you design a research project like this, you have to be very limiting in terms of qualifying participants?
Yes, you should be constricting that as much as possible, because including many covariates in a statistical model can blur the analysis. It’s better to avoid them altogether.
That of course reduces the numbers of subjects who can qualify for the study.
It comes at the expense of reduced numbers and lower generalizability of your results. But you want to limit your biases as much as possible. Because there are so many potentially confounding variables, tons of studies nowadays try every single thing you can imagine, and as a consequence you can conceptualize almost any result and find a paper that reports it, so it’s a bit of a mess.
Have you found any surprises in your research on traumatic brain injury?
During my time as a graduate student and postdoc, we published four papers on TBI —doing different post-processing and different ways to detect injury in one single cohort. The idea was to get at the best way to use spectroscopy to detect injury and at MRS’s potential as a biomarker. We investigated grey matter, white matter, cortical structures, different regions of white matter, correlations with clinical outcomes, looking for biomarker properties. But there’s a tremendous amount of literature that didn’t jibe with our conclusions, so I always felt a nagging question about how generalizable the results are.
And that reflects the environment you described, in which there’s a lot of information pointing in different directions.
Yes, and so, recently, together with Anna Chen who is a PhD candidate here, we recruited a very similar cohort, and I gave her the task to see whether what we had found in those four papers replicates.
We basically rolled the experiments we had done in those papers into one study, testing each hypothesis resulting from the earlier works. And, shockingly, most findings replicated. I was stunned.
What did that reproducibility mean to you? Was that a big confidence boost in your line of inquiry?
It was, because for the original cohort we came up with a technique that can detect injury on a global scale, on the scale of the entire brain. If something is happening only in a small portion of the brain, it would be left undetected, but if there’s a chemical imbalance across the brain, this would be a great way to look at it.
The advantage of looking at the whole brain is that you use all your signal, all your data. This matters in spectroscopy because signal from metabolites is fifty thousand times weaker than from water. And what we found in the first cohort gave us very prominent differences between patients and controls, and gave us correlations with clinical outcomes, which are the two things you want from a biomarker. And we found this exact same thing in the second cohort—the only difference was that different metabolites appeared as abnormal. So, the actual injury in terms of chemical imbalance was different, but the injury distribution was the same.
After a finding like this, what is the next step?
The fact that it replicates strengthens the case for pursuing further inquiry into assessing MRS as a potential clinical biomarker of TBI, but we need to conduct further experiments to get to that.
You’re leading an NIH-supported investigation that is a collaboration with NYU Langone’s Alzheimer’s Disease Research Center. Can you talk about that study?
I’m co-principal investigator on that study—the other co-PI is Assaf Tal. We’re using something that’s not conventional MRS; it’s MRS fingerprinting, a sort of cousin of magnetic resonance fingerprinting, aimed at acquiring relaxation times of both water and metabolites simultaneously. It’s a new angle that allows us to measure the relaxation times of chemicals very quickly.
The fingerprint principle is that you can match signal to a dictionary, and that’s where the speed comes from.
Exactly. And the novelty is that you can have the relaxation times of metabolites in five minutes as opposed to fifty. These carry unique information because they come from molecules that are cell-specific. You get the T1 and T2 relaxation times of NAA, and NAA resides only within neurons, so you actually have a marker for the molecular environment within neurons—and, if you want to be even more specific, for the viscosity of that environment.
And what is viscosity a proxy for?
It’s a proxy for macromolecular concentration within the cell and for structural crowding, meaning neurofilaments and microtubules, the structural elements of the cell.
Why is viscosity so interesting in a condition like Alzheimer’s?
For one, because very high-profile research links the accumulation of tau protein to increases in viscosity within the neurons before these accumulations become the insoluble tangles that you see on positron emission tomography.
So, viscosity could be a very early signal of potential Alzheimer’s.
Exactly. I’m really excited about the biology behind it. That’s where my background as a neuroscientist, as someone whose PhD exam was on basic neuroscience and physiology, gets me excited about the imaging, because I can connect the biology to the new imaging technologies and look for where they can be helpful to people.
You earned your PhD and completed postdoctoral training here at NYU Grossman School of Medicine. How did you find your way to PhD studies in the first place?
I’ll start from very early on. I come from four generations of doctors on my dad’s side, and my grandfather on my mom’s side was also a physician—a pulmonologist in Sofia, Bulgaria, where I grew up. So, there was always a lot of talk about how it would be appropriate for me to be a fifth-generation doctor, and helping people sounded amazing to me. When I was a high school junior, I moved with my dad to California. My dad—a physician—basically started over in the United States, taking exams, doing residency, fellowship, and then rising through the ranks. He was a director of a hospital cancer institute when he retired.
I was inspired by him and wanted to see whether medicine was for me. But in college I did some shadowing with him and some volunteering in an emergency room, and the bottom line is that I didn’t like blood. I would get the fainting reflex or feel nausea at the sight of the large syringes for drawing cerebrospinal fluid and things like that. I was squeamish about it.
You learned in a visceral way that medicine wasn’t in the cards for you.
And so, a PhD sounded appealing: venturing into the unknown and also having broader possibilities than just to be a doctor.
It sounds like it was a path that allowed you to stay within the realm of your interests and values but also opened for you new territory for professional activity.
It also fit my natural strength, which is in the details, in doing things methodically and wanting to thoroughly understand them. In college, I was interested in studying the origins of life, basically how microbes can survive in extreme conditions and how you can build higher organisms from lower organisms. But then I took a psychology class and really loved it, and that led me to a class in neurobiology. There was a statement in the textbook that said we know almost nothing about the brain, and yet this is what drives us, it’s what we are—that was a moment of scientific fascination for me.
You weren’t thinking small, picking between the biggest mystery of human physiology and the mystery of life itself.
That’s how young brains work—I was a different person at that point, a person whose life is ahead of him and who is always boggled by these questions. So, those were my main inspirations to do a PhD.
You’re now a member of the mentoring faculty in the PhD program you yourself came up through. At the start of the program, students do three lab rotations before deciding what to focus on. What were your rotations?
I was awarded a physiology and neuroscience training grant, and my first rotation was in a wet lab. My second rotation was with Oded Gonen, and the rest is history.
Dr. Gonen—now professor emeritus of radiology at NYU Langone—is an expert in the development and application of MRS methods to the study of the human brain. So, in his lab you came across a method directly relevant to your major area of interest.
And also a way to get closer to patients and to a clinical aspect of research that can benefit them—without the syringes.
We have talked about your investigations of TBI and Alzheimer’s, but haven’t touched on multiple sclerosis. Is there anything you’d like to say about your research on this particular condition?
I’m especially proud of one paper on MS. It’s a study of lesions, done with semiannual follow-ups over the course of three years to characterize lesions from A to Z. It was a highly comprehensive work describing what the lesions look like on MRI and what their chemistry is on MRS at each stage. And because some of the lesions developed over the course of the study, we were able to go back and see what the tissue looked like six months before they showed up on MRI. We showed that the chemistry of pre-lesional tissue is abnormal compared to the tissue around it, and that in lesions that resolved—because a lesion can also disappear—their NAA does recover.
So you found that in those lesions that go away, the tissue doesn’t just look normal on an MRI but its neurons also recover.
Yes, and that’s what people use in clinical trials—how many lesions stay or disappear? And the ones that disappear, you can actually confirm that the tissue recovers also from a metabolic point of view.
What makes you especially proud of this study?
What I’m proud of is that in our studies, we try to be as rigorous as possible, to be as confident as we can be with the data available; we try to show the data and be transparent, and to work with hypotheses that are well-defined and well-supported.
I’m really proud of the rigor of what we do in the lab, and I try to teach my students to do solid hypothesis-based science. That’s what really makes me happy in my work. I prefer to have big papers that are comprehensive and well done rather than little bits here and there.
Related Story
Anna Chen, PhD candidate in biomedical imaging, talks about what led her to research, why she's working with MR spectroscopy, and how she's learning about business.