Single-value brain activity scores reflect both severity and risk across the Alzheimer’s continuum

Abstract

Single-value scores reflecting the deviation from (FADE score) or similarity with (SAME score) prototypical novelty-related and memory-related functional magnetic resonance imaging (fMRI) activation patterns in young adults have been proposed as imaging biomarkers of healthy neurocognitive aging. Here, we tested the utility of these scores as potential diagnostic and prognostic markers in Alzheimer’s disease (AD) and risk states like mild cognitive impairment (MCI) or subjective cognitive decline (SCD). To this end, we analyzed subsequent memory fMRI data from individuals with SCD, MCI, and AD dementia as well as healthy controls (HC) and first-degree relatives of AD dementia patients (AD-rel) who participated in the multi-center DELCODE study (N = 468). Based on the individual participants’ whole-brain fMRI novelty and subsequent memory responses, we calculated the FADE and SAME scores and assessed their association with AD risk stage, neuropsychological test scores, CSF amyloid positivity, and ApoE genotype. Memory-based FADE and SAME scores showed a considerably larger deviation from a reference sample of young adults in the MCI and AD dementia groups compared to HC, SCD and AD-rel. In addition, novelty-based scores significantly differed between the MCI and AD dementia groups. Across the entire sample, single-value scores correlated with neuropsychological test performance. The novelty-based SAME score further differed between Aβ-positive and Aβ-negative individuals in SCD and AD-rel, and between ApoE ε4 carriers and non-carriers in AD-rel. Hence, FADE and SAME scores are associated with both cognitive performance and individual risk factors for AD. Their potential utility as diagnostic and prognostic biomarkers warrants further exploration, particularly in individuals with SCD and healthy relatives of AD dementia patients.

Reduced expression of fMRI subsequent memory effects with increasing severity across the Alzheimer’s disease risk spectrum

“…To this end, we have employed contrast maps obtained with the winning theoretical parametric GLM (i.e., the model using the arcsine-transformed memory regressor) to calculate single-value scores (Soch et al, 2021b; Richter et al, 2023). In a direct follow-up to the present study, we describe the extent to which these scores can further differentiate between the diagnostic groups in the clinical sample described here (Soch et al, 2024). Furthermore, we are currently working on improved computational modeling of the subsequent memory reports, which could be used in the future to differentiate participant groups based on purely behavioral response patterns (Soch et al, 2022).…”

“…This suggests that, with reduced behavioral accuracy, the predictive value of neural signals with respect to later remembering and forgetting also drops, such that fMRI responses do not covary with subsequent memory reports when the latter most likely reflect mere guessing, at least to a considerable proportion. On the other hand, the declining subsequent memory effect across the AD spectrum could also be interpreted as indicative of the memory decline, with a potential diagnostic or prognostic utility (Soch et al, 2024).…”

“…Compatibly, individuals with SCD and healthy relatives also showed behavioral memory performance and fMRI signals comparable to that of healthy controls (see Figure 1 and 6). This points out the need to stratify SCD into those with subjective complaints and normal performance vs. those with subjective complaints and sub-normal performance (Koppara et al, 2015), possibly based on amyloid pathology (Jessen et al, 2022; Soch et al, 2024). Like the SCD group, healthy relatives often showed model preferences and activity patterns qualitatively identical to those of HCs (see e.g., Figures 2, 3 and 6), consistent with AD relatives in the DELCODE study not significantly differing from HCs in terms of MMSE total (see Table 1), NPT global score, the PACC5 neuropsychological composite score, or ApoE genotype (see Soch et al, 2023b).…”

“…However, in our study, we addressed specific contrasts rather than BOLD signal relative to baseline. Furthermore, differences between diagnostic groups do not only include reduced deactivations, but also atypial activations on the memory contrast (i.e., effects of subsequent memory) in DMN regions for Alzheimer’s disease patients (see Figure 1E in Soch et al, 2024). Therefore, we conclude that, although the potential impact of cerebrovascular differences cannot be excluded, it is, in our view, unlikely that vascular effects are the main drivers of our results.…”

“…On the other hand, the declining subsequent memory effect across the AD spectrum could also be interpreted as indicative of the memory decline, with a potential diagnostic or prognostic utility (Soch et al, 2024).…”