2026 American Academy of Neurology Abstract Website

Steph Maynez¹, Annabelle Shanks¹, Hailey Brigger¹, Ian Johnson¹, Alison Champagne¹, Rachel Hird¹, Gordon Sze², Jua Iglesias Gonzalez³, Adam de Havenon¹, Kevin Sheth¹
¹Department of Neurology, Center for Brain & Mind Health, ²Department of Radiology, Yale New Haven Hospital and Yale School of Medicine, ³Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology

Objective:

To quantify the impact of successive software versions on hippocampal volumetric reliability in portable low-field MRI (pMRI), relative to conventional MRI (cMRI).

Background:

pMRI systems operating at 0.064 Tesla have undergone multiple software updates that may influence the accuracy of hippocampal volumetric measurements, a quantitative marker of neurodegenerative progression and postoperative recovery outcomes. Given these software updates, evaluating their impact on volumetric consistency is essential for validating pMRI in clinical contexts.

Design/Methods:

We analyzed 178 paired pMRI and cMRI scans obtained between 2018 and 2022, each acquired within 24 hours of one another using FLAIR sequences. Hippocampal volumes were segmented automatically using WMH-SynthSeg, and results were grouped by software version (8.1, 8.2, 8.3). Percent differences in hippocampal volumes between pMRI and cMRI, Pearson correlation coefficients, and one-way ANOVA were utilized to evaluate volumetric reliability across versions.

Results:

Across software versions, mean hippocampal bias between pMRI and cMRI remained consistent (F(2,175)=0.82, p=0.44): –8.0 ± 18.1%, –8.8 ± 14.0%, and –11.3 ± 10.4% for versions 8.1, 8.2, and 8.3, respectively, with pMRI consistently underestimating hippocampal volumes by roughly 10%. Variability of the pMRI-cMRI volume differences decreased substantially across software versions (8.1: 328.79 mm³; 8.2: 197.18 mm³; 8.3: 107.28 mm³). There was no significant correlation between hippocampal volumes measured on pMRI and cMRI in version 8.1 (r = -0.17, p > 0.05), while strong positive correlations emerged in versions 8.2 (r = 0.62, p < 0.001) and 8.3 (r = 0.59, p < 0.001), demonstrating improved agreement between pMRI and cMRI measurements in newer software versions.

Conclusions:

Successive pMRI software updates substantially improved hippocampal volumetric reliability relative to cMRI. These findings underscore that continued software refinement enhances the quantitative accuracy of portable low-field MRI, advancing its clinical applicability for monitoring neurodegenerative changes.