Normative morphometrics from MRI are associated with episodic memory test performance in amnestic MCI

Background Proprietary software that generate normative scores for brain regions using MRI (e.g. NeuroQuant, Neuroreader) have identified greater atrophy in MCI patients than controls, particularly in the hippocampus, and have associated these measurements with episodic memory performance (Tanpitukpongse et al., 2017; Umfleet et al., 2020). While useful in clinical settings, these may be cost‐prohibitive for research purposes. Our goal was to conduct a validation study using free, non‐proprietary morphometric software (Potvin et al., 2016; 2017) with data from a cross‐sectional study of amnestic MCI (aMCI) and normal controls (NC). Method 102 older adults ages 60 to 86 (n NC =75; n aMCI =27) underwent 3T brain MRI and neuropsychological testing (Weintraub et al., 2018). T1‐weighted images were segmented using FreeSurfer v5.3 , and from these outputs normed z‐scores were generated, adjusting for age, sex, total intracranial volume, and MRI field strength (Potvin et al., 2016a; 2016b; 2017). We hypothesized that volumes of AD‐relevant regions (hippocampus, amygdala, entorhinal cortex, parahippocampus) would be smaller in aMCI than NC with no differences in non‐AD “control regions” (cuneus, lingual gyrus, pericalcarine; Fig. 1.A), and that AD‐relevant volumes would correlate with episodic memory composites (i.e. average of normed scores (percentiles) on immediate, delay, and recognition trials from list‐learning (Rey‐Auditory Verbal Learning Test) and story‐learning (Wechsler‐Logical Memory) tests). Result Bonferroni‐corrected t‐tests showed significantly lower volumes of all AD‐relevant regions in the aMCI versus NC, with large effect sizes (Hedge’s g =0.9 to 1.2), and no group differences in control regions ( p s<0.05, Hedge’s g <0.3). Bivariate correlations showed moderate ( r s = 0.38 to 0.55) associations between memory composites and most AD‐relevant regions in the aMCI but not the NC group (Fig. 1.B.1), and no associations between memory composites and control regions except for a positive association between story‐learning and cuneus volume in aMCI (Fig. 1.B.2). Conclusion Similar to results using proprietary software, this study validates the potential clinical utility of a non‐proprietary normative morphometric software when distinguishing aMCI from NC, with AD‐relevant volumes associated with episodic memory in aMCI. Future studies are needed to characterize associations between other AD‐relevant brain regions and cognitive domains.