MR Fingerprinting: A Promising New Approach

Podcast Transcript

Intro: Neuro Pathways, a Cleveland Clinic podcast, exploring the latest research discoveries and clinical advances in the fields of neurology, neurosurgery, neuro rehab, and psychiatry.

Glen Stevens, DO, PhD: Magnetic resonance fingerprinting has emerged as an alternate imaging method that can successfully address problems associated with conventional and advanced imaging modalities. In today's episode of Neuro Pathways, we're discussing MR fingerprinting and its promising new approach to standardized imaging biomarkers. I'm your host, Glen Stevens, Neurologist, Neuro-oncologist in Cleveland Clinic's Neurological Institute. I'm very pleased to have doctors Daniel Ontaneda and Irene Wang join me for today's conversation. Dr. Ontaneda is a Staff Neurologist in Cleveland Clinic's Mellen Center for Multiple Sclerosis, and Dr. Wang is the Research Director and Staff Scientist in the Charles Shor Epilepsy Center in Cleveland Clinic's Neurological Institute. Dan and Irene, welcome to Neuro Pathways.

Irene Wang, PhD:Thank you, Glen. Happy to be here.

Daniel Ontaneda, MD, PhD: Thanks so much, Glen.

Glen Stevens, DO, PhD: In 1919, Walter Dandy invented the pneumoencephalogram. 100 years later, look at where we are now to image what's going on in the brain. I must admit that I don't know much about this technique, so I'm glad content experts like yourself are here today. My understanding of MRI fingerprinting is that it's not so much about the imaging, but it's about data. So, it sounds like we're going to get our geek on here today, and we'll start with Dr. Wang. MRI fingerprinting is an emerging quantitative MRI technique that is being used in clinical research. Can you start off today's conversation and tell us in simplest terms, what is magnetic resonance fingerprinting?

Irene Wang, PhD: MR fingerprinting, or MRF, is a novel and flexible MR acquisition framework that allows for fast acquisition of multiple quantitative tissue maps. This technique can be performed on a conventional clinical MRI scanner, and can be used for many applications of brain and body MR imaging. MRF was originally invented by our colleagues at Case Western Reserve University, Dr. Mark Griswold, Dr. Dan Ma and the group. The first MRF paper was published in Nature 2013, and continued research in this area has created a whole new direction in the field of MRI. So, how does it work? MRF operates on the premise that each tissue has its own distinct fingerprint that can be identified through a new approach to acquire the MRIs. So, instead of doing serial acquisition of different sequences, MRF uses a randomized acquisition with varying values for image parameters, such as the flip angle, repetition time, echo time, and inversion time so that different tissues have different signal evolutions.

The result of the randomized acquisition is a set of time resolved images and the time course for each pixel or voxel reflects his underlying tissue properties. And ahead of time, a dictionary has been already created by simulating all the signal time courses that could possibly appear based on a whole wide range of physiologically possible tissue property values. The difference signal evolutions from the acquisition is then matched to find the dictionary entry that most resembles its characteristics, and this is done for each pixel or each voxel. And from that, we can get the information we're interested in T1, T2, and so on, that made up the signal evolutions. So, in this way, the MRF technology enables the noninvasive quantification of multiple properties of tissue simultaneously.

Glen Stevens, DO, PhD: Is it done on a 1.5, a 3.0, a higher strength magnet, lower strength magnet? Does it matter?

Irene Wang, PhD: It has been worked out that it can be done on 1.5T or a CT, and on the Ultra High field 7Tesla MRI still there's a lot of work to be done to make it possible.

Glen Stevens, DO, PhD: Great. So, we'll move over to Dan. Compared to conventional magnetic resonance imaging, can you describe the specific advantages of MRF? And you can discuss quantitative advantages, speed, harmonization, reproducibility.

Daniel Ontaneda, MD, PhD: Yeah, I think Irene made a quite clear description of what the MRF technique is. And many of those qualities translate into what the benefits of using MRF are for us as researchers. And then we hope eventually in clinical practice. And I think the first and perhaps most clear advantage is the fact that MRF provides quantitative measures. Something that people might not realize is that the majority of the MRI images that we look at on a day-to-day practice aren't actually T1 or T2 relaxation values. What they are is something we call weighted images and they're basically weighted on a gray scale, but they don't really mean anything in terms of a numerical or a quantitative value. The difference with MRF is that MRF can produce this quantitative values from the scanner, and that will have significant advantages in terms of harmonizing what an MRF image acquired looks like or is measured like in Cleveland, for example, to one measure in Boston or say in San Francisco.

The idea here is that if you've got a quantitative number, that won't mean the same irrespective of where you are, and that's not the truth for a weighted image on MRI where image acquired on an MRI in a different location will be significantly different.

I think one of the other advantages is that the way we model MRF, it allows us to characterize tissue even in areas smaller than a single voxel. In MRI, typically what we have in weighted images is that that voxel, that unit area that we're looking at, if that's in boundary tissue, for example, it will be either white matter or gray matter. And if it's in boundary, it will be a combination of both. And that's a problem that we call partial volume averaging that occurs with MRI that is in a single box. When you have two components, you don't know where the actual border is, if it's in that box or it's out. Now, what MRF allows us to do is allows us to derive, for example, measures that we might think are coming from white matter, measures that we're thinking are coming from gray matter.

And we can say in a given voxel, what percentage of that voxel is white and what percentage of that box is gray therefore solving this problem of partial volume averaging, which is something that has really marred MRI for years. Reproducibility, we talked a little bit about this already, which is, the numbers we generate on MRF are going to be reproducible within a given patient. So, if I scan you today, Glenn, and then I scan you three days later, your numbers, if nothing's happened neurologically to your brain, will look pretty similar. And the idea is that, if we're trying to compare groups of patients across different sites, they will be analyzable as basically on the same scale on the same metric. And this has significant implications for the application of MRI, for example, in multicenter studies. One of the other advantages of MRF is robustness this to motion.

And this is something that if anybody's had an MRI knows, we tell patients, you have to be as quiet and as still as possible while you're in the scanner. And this is because small movements actually cause distortion on images. Now, because MRF is based on a randomized sequence, that randomized sequence is under sampled and therefore, it's a technique that is robust to motion. And so the nature paper that Dr. Wang described, what they had patients do is tilt their heads 15 degrees while they were in the scanner. And despite that fact, the images generated from those acquisitions with motion basically could be reconstructed with full fidelity. And then the final thing I think is speed of acquisition. Normally, when you acquire MRI, you have to acquire each parameter separately. So, you do a parameter for T1, a parameter for T2, a parameter for, for example, on diffusion weighted imaging or something else.

With MRF, one of the propositions is that you can gather multiple contrasts in a single scan. And now, we're down to the point that we can acquire full brain imaging in three to five minutes using MRF with 3d applications, basically accelerating the speed and the convenience for patients.

Glen Stevens, DO, PhD: Excellent. You guys are so good. I'm starting to understand this. And I'm glad that you mentioned the acquisition time because I'm shocked at how short the acquisition time is with that. I mean certainly dramatically different than a standard MRI. So, let's move to your respective diseases of interest and, Dr. Wang, we'll start with you first in the epilepsy area. Could you discuss a use of quantitative maps to inform lesion detection, which I know is a big thing in epilepsy and disease characterization?

Irene Wang, PhD: Sure. So, I would like to echo what Dan just very nicely summarized about the advantages of MRF and start by discussing some of the work we do for epilepsy. So, the importance of MRI cannot be overstated for the diagnosis and surgical treatment for epilepsy. In the process of pre-surgical evaluation, which we do a lot in our epilepsy center, we always strive for finding the lesions that cause the epilepsy. Although clinical MRI can already see a fair percentage of epileptic lesions, we still have about one third of patients who undergo evaluation for epilepsy surgery, but had a negative or non-lesional MRI. It's very challenging to manage this group of patients. And this really calls for improvement of MRI techniques with increased sensitivity. So, a few years ago, we started a close collaboration from the Cleveland Clinic Epilepsy Center with Dr. Dan Ma and Dr. Mark Griswold from Case Western Reserve University and Dr. Stephen Jones from Cleveland Clinic's Imaging Institute to explore the potential of MRF in clinical application for epilepsy.

We were later fortunately awarded a five-year NIH RO1, to support this exciting project. So, what are the key areas to look at? With the additional information from the T1, T2 quantitative tissue maps, are we able to see more lesions, especially the subtle ones that we missed on the conventional MRIs? Are we able to better characterize visions, for example, no, the subtype of focal cortical dysplasia before going to surgery? And when they're multiple lesions, can we better be informed which lesions are more relevant to the epilepsy? And a more paradigm shifting question is, can a 10 minute 3D MRI sequence replace the 30 minutes clinical MRI protocol and save hours and hours of scanner time while still containing the same clinical information? So, we're in the process of figuring out questions to these questions and there have already been exciting results coming out.

We published the first study on the pilot cohort of 15 patients on the journal of MRI. In four of the 15 included patients who had varying types of epileptic lesions so this is quite a heterogeneous group, MRF was able to show additional clinically relevant information, not revealed by conventional MRI. So, we are continuing our patient recruitment and technical development to see if these results can be further replicated in bigger cohort studies. We have also recently started to go outside of the box of lesion detection, but additionally, leverage the quantitative nature of MRF to evaluate or characterize the influence of epilepsy on the whole brain. It has to be more and more recognized that epilepsy is a progressive disorder. When the brain has years and years of seizures, can we use MRF to measure the structural changes happening in the brain that may reflect the disease progression? So, all these directions are very exciting and all our efforts hopefully contribute to applying MRF in the field of epilepsy as a novel tool to examine the intrinsic properties of epileptic pathologies at a final finer level than ever previously before

Glen Stevens, DO, PhD: Have any of those patients gone to surgical resection, and if they have, what did the pathology show?

Irene Wang, PhD: Yes, I can give you some specific examples. But, yes, one of the very nice things to validate MRIF in the field of epilepsy is that we always have pathology confirmation of the results. So, if we see something that we were not able to see on the clinical MRI, we are always able to validate what it is in the surgical specimen. So, yes, from the pilot study that we published on the four patient out of the 15 who had additional information, those were confirmed by surgical pathology.

Glen Stevens, DO, PhD: Okay. Good. So, we'll move back over to Dan in the field of multiple sclerosis. And I wonder if this is the area of be careful what you ask for, and maybe you'll see too many lesions and then you'll have to come to justification of how you manage that. But talk to us a little bit about the use of MRF in multiple sclerosis, Dan, clinically.

Daniel Ontaneda, MD, PhD: So, similarly, MRI is a time-tested method that we use in multiple sclerosis both to make a diagnosis and also to follow treatment response. And we do this mainly by identification of white matter lesions both in the brain and in the spinal cord. But we've known for years that despite how useful MRI is in MS, there is something called a clinical radiological paradox. That is, some patients have MRIs that don't look that bad, but clinically they're doing not well at all. And we have other patients whose MRIs look terrible. They have a lot of lesions and you would expect them not to be well clinically, but they're actually doing okay. And we also know based on our pathological study, several of them done at the Cleveland Clinic, that there's a lot of heterogeneity between one MS plaque and another MS plaque. We know that about 30% of lesions that we find on MRI in MS, actually aren't even demyelinated.

We published a couple of years ago, we think about 10% of patients from our post-mortem program who had no cerebral demyelination whatsoever, but their MRIs looked like MS. Otherwise, what this leads us to conclude is that perhaps MRI is sensitive for diagnosis of MS and is sensitive to demyelination, perhaps it's not specific and, similar to what Dr. Wang described in epilepsy, we know that MS also has a neurodegenerative component. That is, there is a process in the brain that works in slowly over time, makes the brain shrink. And so we've adapted the use of MRF specifically to answer questions related to these issues. And we've done two studies. One was a study that was conducted among 55 subjects. Some of them had MS. Some of them were controls and what we try to examine is, can we find a fingerprint or a signature that lets us diagnose MS based on what their MRI data is showing us. And indeed, we found that there were significant differences between healthy controls compared to the early MS patients compared to later MS patients. We have reproduced that data in a second study, where we specifically were focusing on a deep structure in the brain called the thalamus.

The thalamus, if you'd like to think of it, is kind of grand central station for the brain, everything that comes in comes out of the thalamus. It connects with all brain regions and really processes all the incoming and outgoing of flow of information. And we know that the thalamus is one of the areas that changes earliest in MS. It actually starts shrinking, and we know that it contains both white matter and gray matter. It has both lesions in it, that is plaques and also has non-lesional pathology. So, we thought that the thalamus was probably the best place to use a sequence like MRF to study the micro structure of that organ and see, well, what is actually changing. And what we found in those studies, which was quite interesting, was that the thalamus was actually changing, not because of changes in the relaxation values in the thalamus itself, but because of changes in the relaxation values in the white matter outside of the thalamus. It was somewhat surprising, I have to say, we were expecting to find a signature within the thalamus, and what we found is that the shrinkage of the thalamus is actually responding to tissue degeneration outside the thalamus. So, instead of thinking of the thalamus as a generator for neuro degeneration, we think of it more as a barometer of neuro degeneration.

Future studies, I think, are going to look at, can we find a specific fingerprint either in white matter or in gray matter that might indicate MS as a diagnostic tool. And the other thing we want to do is try to validate MRF as perhaps something that can show us or determine early on if a patient has actually responded to treatment or not. All these issues will need more perspective studies and prospective data collection. But because the sequence is easy to acquire, it's short, it's an ideal for applications related to studies that are done in multiple sites across the country.

Glen Stevens, DO, PhD: Excellent. Thank you, Dan. Irene, I'll go back to you. Are there any other examples of lesion detection for epilepsy surgery that you'd like to share with us in your experience?

Irene Wang, PhD: I would like to first clarify that so far in epilepsy, we have been performing MRF as IRB approved research study. Although the MRF maps are not yet directly used for clinical practice due to its investigational nature, the findings of MRF are compared in parallel with multimodal findings from standard of care, as well as surgical location and surgical pathology as Glen you asked earlier and seizure outcomes. So, I'll give you a couple of examples that may highlight the additional value MRF can generate. The first example is for periventricular nodular heterotopia, which is a type of cortical malformation frequently associated with medically intractable seizures. This type of malformation usually present with multiple nodules along the ventricles, sometimes bilateral and on a clinical MRI scan, the lesions appear to have the same signal intensity. So, you're not able to distinguish one from another. We have reported a very interesting case where MRF tissue property maps, some nodules were significantly different from the rest of them.

This signal intensity difference was never seen on the conventional MRI. So immediately asked, is this relevant to the epilepsy? After invasive evaluation with intracranial EEG, these nodules with a distinct signal differences were indeed confirmed to cause the patient's epilepsy and surgical resection, including these nodules gave the patients seizure freedom. So, this is a great example to show the improved characterization of epileptic lesions. Another example is on lesion detection for subtle focal cortical dysplasia. This is another critical malformation, very frequently associated with epilepsy. Some of these patients are so small, so subtle that they are missed in the reading of clinical MRI. We have reported an intriguing case where MRF maps visualized additional tissue alterations at the deaths of sulcus. so, again, we asked, is this relevant to the epilepsy? Well, we saw EEG monitoring results were consistent with the location of the abnormality and receptive surgery completely included the abnormality and the patient became seizure-free and surgical pathology also confirmed the abnormality. So, this is a great example of the improved detection of epileptic lesions.

Glen Stevens, DO, PhD: Excellent. So, I'll open this up to either one of you. Are either of you aware of other areas in neurology where this technique is being used or where it could be used, but isn't being used?

Daniel Ontaneda, MD, PhD: Yeah, I think one of the classical examples where MRIF has been found to be useful is characterization of neuro-oncology so I feel that you're very familiar with, Glenn. And preliminary studies conducted at Case Western, essentially, were able to use MRF and the fingerprint generated from different tumors to differentiate, for example, primary brain tumors from metastatic brain lesions. And that was an evolution of work that has also been done. It's been done not only in brain tissue, but it's also been done to differentiate, for example, benign from non-benign disease in the prostate, benign from non-benign disease in the liver, in the applications of MR body approaches. But I think that characterization, for example, of conditions that produce degeneration of the brain with no focal tissue is probably one of the areas that MRF has to move into. And the main thing that, I think about going forward, is perhaps neurodegenerative diseases, including dementia.

Glen Stevens, DO, PhD: Irene, anything else to add?

Irene Wang, PhD: Yeah, another thing I would like to add is that because MRF is a new technique, post-processing of the MRF images still needs a lot of further development. This is because existing image processing software were developed for conventional MR images and do not work just out of the box for MRF images. So, the task of our team now is actually to combine the novel MRF technique with post-processing analysis and machine learning, deep learning algorithm. This is a strategy that we have already used successfully with conventional MRI, and we'll continue to use it in combination with a novel MRI techniques. The combination of the innovations will really allow for unleashing the power of MRI for epilepsy clinical applications.

Glen Stevens, DO, PhD: Well, Dan and Irene, thank you so much for joining me today. This has been very insightful conversation, and I appreciate your time.

Irene Wang, PhD: Thanks for having me.

Daniel Ontaneda, MD, PhD: Thanks, Glen.

Outro: This concludes this episode of Neuro Pathways. You can find additional podcast episodes on our website, ClevelandClinic.org/neuropodcast or subscribe to the podcast on iTunes, Google Play, Spotify, or wherever you get your podcasts. And don't forget you can access real-time updates from experts in Cleveland Clinic's Neurological Institute on our consult QD website. That's consultqd.clevelandclinic.org/neuro or follow us on Twitter @CleClinicMD, all one word and thank you for listening.