Abstract
Recent research has found that certain images over others are particularly diagnostic of memory impairment in Alzheimer's Disease (AD), where these images are better at predicting whether a person has a mild cognitive impairment (MCI) or is a healthy control (HC) (Bainbridge et al., 2019). However, it is unclear why these images cause such group differences, and how these behavioral differences might relate to biomarker and neuroimaging data. In the current study, we explored whether these images indicate biomarker differences and elicit different fMRI responses for HCs versus MCI individuals, analyzing data from the large-scale DELCODE project (Jessen et al., 2018). 64 HCs and 48 MCI individuals viewed scene images in the scanner and were later tested on the recognition memory of these images. Tau and amyloid biomarkers (total tau, phospho-tau, Aβ42/Aβ40, Aβ42/phospho-tau) in cerebrospinal fluid were also collected. We compared how memory performance for different images was related to the biomarker status of individuals and found that diagnostic images showed higher correlations with these biomarkers than non-diagnostic images. We then analyzed fMRI activation in three regions of interest related to scene perception: the parahippocampal, medial, and occipital place areas (PPA, MPA, OPA). We found a significant interaction between participant group (HC or MCI) and image diagnosticity (diagnostic vs. non-diagnostic) in the left PPA, MPA, and OPA. Specifically, diagnostic images showed higher activation in these regions for MCI individuals, but no difference was found in these regions for HCs. Our results indicate that diagnostic images could better reveal biomarker status and abnormal brain activations related to AD pathologies than other images, pointing out the potential of improving cognitive assessment with these images.