NeuroVoices: Irene Wang, PhD, on the Clinical Utility of MR Fingerprinting in Epilepsy

SAP Partner | <b>Cleveland Clinic</b>

The research director and staff scientist at Cleveland Clinic’s Epilepsy Center discussed why MR fingerprinting holds significant clinical potential in epilepsy and epilepsy-related disorders.

At the 2021 American Epilepsy Society (AES) annual meeting, December 3-7, in Chicago, Illinois, research by Irene Wang, PhD, and colleagues, adopted a MR fingerprinting-based radiomics feature extraction approach to characterize focal cortical dysplasia (FCD) lesions. A total of 43 healthy patients and 21 patients with epilepsy who had pathological confirmed FCD, either type 1 (n = 9) or type 2 (n = 12), were included in the analysis, for each of whom 3D whole-brain MR fingerprinting was acquired.

A total of 96 radiomic features in the regions of interest (ROI) were extracted from T1, and T2 maps using the radiomics toolbox in MATLAB. The same ROIs were applied to healthy controls to extract average values of features at the matched locations. Wang and colleagues generated 2 logistic regression multivariable models and divided datasets into a training set (80%) and a testing set (20%). In the testing set, the investigators performed backward elimination for the optimal feature selection based on P values between groups.

At the conclusion of the analysis, the MR fingerprinting model with the optimal 30 features showed high area under the curve (AUC), accuracy, sensitivity, and specificity of 0.86 (+0.02), 84% (±1), 87% (±1), and 80% (±2), respectively, in the training set. In the testing set, the model showed stable performance of 0.86 (±0.02), 81% (±2), 86% (±3), and 76% (±4) for the respective values.

In part 1 of an interview with NeurologyLive®, Wang, the research director and staff scientist at the Comprehensive Epilepsy Center of Cleveland Clinic, provided a detailed background on the need to conduct this study, and the origins of MR fingerprinting. She also discussed how this approach is different from other typical standard MRIs, which may be limited in capacity for some forms of epilepsy.

NeurologyLive®:Why are you particularly interested in this MR fingerprinting?

Irene Wang, PhD: This technique was invented a few years ago by our close collaborator, Dr. Dan Ma, PhD, and Dr. Mark Griswold, PhD, from Case Western University. We are at the stage now of utilizing this novel technology to benefit our patients. One of the first areas that MR fingerprinting was applied was on brain imaging and body imaging. In terms of brain imaging, epilepsy was one of the earliest frontiers so to speak. We started collaborating a few years ago and we’re funded by an NIH RO1 project. Since then, we’ve been doing some interesting studies and core collaborations.

MR fingerprinting is quantitative in nature. I like to use this analogy to give people an idea of what this is. Imagine you go home, and your kid is having a bit of a fever. As a mom or dad, you would feel on the kid’s head to see if the fever is high or low. You have a feel and the next thing you do is take out a thermometer to take a measurement of the temperature. Instead, you don’t stop at feeling just high or low temperature, you do an actual measurement. Currently, what we do with MRI is look at dark or bright. This is similar to feeling high or low fever, but what we actually need to do is have that specific number that could tell us the state of the brain, the health state of the brain, so to speak. This is what this quantitative MRI technique can provide us.

Can you provide background on the results you observed when characterizing local FCD lesions?

This is a study that we got a young investigator award; one of our fellows led the study. The results were very interesting. Our group, before this collaboration on MR fingerprinting, have a long history of doing post-processing on clinical MRI images to pull out information that is not so visible to the naked eye. We are now combining this quantitative-MRI with quantitative post-processing. I think of this as quantitative based on quantitative, so its essentially quantitative squared.

What we did in this interesting and exciting study is use a technique called radiomics, which is a post-processing technique, based on the MR fingerprinting input. We did some comparisons, and everything was done on an individual level. We did some comparisons with focal cortical dysplasia, which is a subtle form of lesion that commonly causes epilepsy. Because of the subtle nature of these lesions, it is oftentimes too difficult for the human eyes to detect. With the MR fingerprinting radiomics approach, we did a comparison between normal brain tissue and FCD brain tissue. We were able to show significant differences in the multiple parameters that MR fingerprinting maps were able to generate. It’s interesting to see the differences exist in the multiple parameters and in features from these multiple parameters.

You used a radiomics extraction approach in your study, why was that? Is this something that can be replicated?

This ties well into the clinical practice implication of this line of research. One of the reasons we chose radiomics is because its radar independent. We rely a lot on the expert expertise of our radiologist colleagues, especially when it comes to these very subtle lesions. Experience comes into play in these cases, and we all know that we need time to accumulate these experiences. It would be nice to have a computerized tool that is radar independent that can help us in our everyday practice. That’s the thinking process that led to the study. With this MR fingerprinting-radiomics tool, we are able to generate actual numbers that are independent of cool inputs, these MR fingerprinting maps. And with further development, hopefully we’re able to involve multiple different centers so that our results can be replicated again. This is an area we’re doing continuous research on.

Transcript edited for clarity. For more NeuroVoices, click here.

REFERENCE
Choi JY, Su T, Hu S, et al. MR Fingerprinting radiomics for characterization of focal cortical dysplasia. Presented at AES Annual Meeting; December 3-7, 2021. Poster 3.235