The DeepGRAI Registry in Multiple Sclerosis


Robert Zivadinov, MD, PhD, offers his insight into the findings from the DeepGRAI registry of thalamic volume measurement on multicenter clinical-quality T2-FLAIR images in multiple sclerosis.

Disclaimer: The American Academy of Neurology requested that all attendees remain masked during the Annual Meeting. This interviewee voluntarily removed their mask for this interview.

Robert Zivadinov, MD, PhD: The DeepGRAI registry study is a registry that has been created involving 30 centers in the United States that collected between 30 and 35 patients with MS, retrospectively, so about 1000 patients have been collected in total. These patients had at least 2 MRI scans, which were about 2.5 years apart. The idea behind this study was to measure thalamic volume as an indicator of thalamic pathology. As you may know, thalamic involvement in patients with multiple sclerosis is extremely important, coming very early in the disease. Even patients with clinically isolated syndrome already have substantial thalamic loss. And in all disease stages, loss of thalamic volume has been shown to be predictive of further progression, and it's constant over time.

The reason why we started this study is to show whether, in the real world, we can measure thalamic volume and determine whether this loss of the thalamus in a nonstandardized setting can be predictive of disability progression. I think the first takeaway is that prior to this DeepGRAI registry study, we needed to create this deep gray matter via artificial intelligence tool, which is an artificial intelligence tool to measure thalamic volume on low-resolution MRI scans that are present in clinical practice and clinical routine. And that's the T2 FLAIR scan. That has been tested validated and published in Neuroimage: Clinical in 2021. Using this tool, we were able to, in a limited setting of the first study, show that it's very robust in measuring thalamic volume and that it can predict disability and cognition.

In this study, by collecting the scans from 30 centers instead of from just one center, we wanted to have a heterogeneity variability of data. More than 33 scanners have been used in this study, with no prior standardization, which means we didn't go and talk to these sites about how to do the scanning in any way, we just took the data that they had on their patients and process it with this tool.

The first finding was that the tool was extremely robust. Only 3.2% of patients failed the analysis. So that means of 1000 people, only 30 patients were not possible to be calculated for thalamic volume. That's extremely important for clinical routine because you want your measure to be applied to every patient. The second thing was that when we looked longitudinally at how much of the thalamic volume change occurred, we found that in patients who progressed from a disability point of view, there was about 1.35% of loss of the thalamus. In patients who remained stable or who improved, it was less than half of that, around 0.5% to 0.7%. Because of that, thalamic volume was an extremely strong predictor of disability progression in this study. This is important because thalamic volume loss represents gray matter damage in multiple sclerosis and it is now possible to have this done on every clinical patient. It clearly gives providers a tool to better measure what's happening from a gray matter point of view.

I would like to say that we also looked, in the study, at how this changes because 43% of the people over the follow-up changed scanners, which affected the thalamic measurement. But despite the scanner changes, we were still able to show good predictability toward disability progression. To some extent, this measure can be done in clinical routine and it's also robust to the scan changes. As for higher-resolution scans, like 3D T1, only 30% of the patients had 3D T1 in that study, and about 60% had 2D T1 in this study. More than half of patients with 2D T1 and more than 70% with 3D T1 did not have thalamic volume produced because the sequences were not available.

Transcript edited for clarity.

Related Videos
Charbel Moussa, MBBS, PhD
4 KOLs are featured in this series.
4 KOLs are featured in this series.
4 KOLs are featured in this series.
5 KOLs are featured in this series.
5 KOLs are featured in this series.
Debra Miller
© 2024 MJH Life Sciences

All rights reserved.