Network-based amyloid-β pathology predicts subsequent cognitive decline in cognitively normal older adults

Abstract

The deposition of amyloid-β (Aβ) protein in the human brain is a hallmark of Alzheimer's disease and is related to cognitive decline. However, the relationship between early Aβ deposition and future cognitive impairment remains poorly understood, particularly concerning its spatial distribution and network-level effects. Here, we employed a cross-validated machine learning approach and investigated whether integrating subject-specific brain connectome information with Aβ burden measures improves predictive validity for subsequent cognitive decline. Baseline regional Aβ pathology measures from positron emission tomography (PET) imaging predicted prospective cognitive decline. Incorporating structural connectome, but not functional connectome, information into the Aβ measures improved predictive performance. We further identified a neuropathological signature pattern linked to future cognitive decline, which was validated in an independent cohort. These findings advance our understanding of how Aβ pathology relates to brain networks and highlight the potential of network-based metrics for Aβ-PET imaging to identify individuals at higher risk of cognitive decline.

Fig. 1:. Cross-validated predictive modeling analysis overview.

The figure illustrates an example of using connectivity-weighted network-level amyloid-β pathology (NAP) with functional-based connectome. a Feature selection on the input of pathology models and connectome. Regional amyloid-β pathology (RAP) or NAP (either connectivity-weighted or centrality-scaled) can be used as input. Individualized network connectomes can be defined based on either structural or functional imaging data. RAP measures do not need the input of connectome information, and the centrality-only model has only connectome information (serving as a benchmark for NAP measures). Since higher pathology is expected to associate with worse cognitive performance, a feature (i.e. regional measures of pathology) is selected if it shows a negative correlation with cognitive change residual score at an uncorrected p < 0.05 in the training set. b Computation of longitudinal cognitive change scores. Subjects are randomly split into a training set and test set. In the training set, we regress out covariate confounds from the longitudinal cognitive change. The covariate-related variability is also removed from the test set using the same weights. c Regression of selected amyloid pathology features in the train set and prediction of cognitive decline in the test set. The sum of selected features is regressed against cognitive change residual scores from the training sample to obtain model weights. The obtained weights are then applied to the features from test set to predict cognitive decline scores, which were then compared to actual scores to evaluate model prediction performance.

Fig. 2:. Regional amyloid-β pathology (RAP) and network-based amyloid-β pathology (NAP).

In addition to RAP, the proposed NAP scores incorporate connectome information by using either a connectivity-weighted or centrality-scaled approach. Here, we show a six-region toy network. a Illustration of RAP deposition $A_i$ in region $i(i=1,\dots ,6)$. b Connectivity-weighted NAP score quantified the influence of Aβ deposition within the connected networks. c Centrality-scaled score quantifies the Aβ deposition in the region and scaled it by its centrality in the connectome. Network connectome information can be generated based on either structural connectome (SC) (d) or functional connectome (FC) (g). d SC is derived from tractography on diffusion tensor imaging data, where SC matrix was denoted as $C_{ij}^S$. e SC based connectivity-weighted NAP score of region $j$ quantifies both the Aβ in region $j$ and a weighed sum of Aβ in all other regions based on connectivity values, where the whole process was denoted as matrix multiplication of $A_i\times \left(C_{ij}^S+I\right)$. f SC based centrality-scaled NAP score of region $i$ is quantified by a Hadamard product of RAP measure and connectivity centrality $A_i\odot \left({\sum }_j^6{C}_{ij}^S\right)$ . g FC is computed from statistical dependency on functional MRI data. h, i For the connectivity-weighted (h) and centrality-scaled pathology (i), FC based NAP scores are computed the same as SC based approach but with FC denoted as $C_{ij}^F$.

Fig. 3:. Performance of cognitive decline predicting performance using different pathology models across cross-validation repetitions.

a Distributions of difference in correlation coefficients between predicted and actual cognitive decline scores across different pathology models. Prediction performance was compared across cross-validation repetitions with regional amyloid-β pathology (RAP) model as the reference. Incorporating structural connectome (SC) with connectivity-weighted network-based amyloid-β pathology (NAP) achieved the highest performance (median R = 0.2682), and outperformed (p < 0.004; test on difference in correlation coefficients against zero) the prediction performance of RAP measures (median R = 0.2013). b Comparison of ratio of average prediction error (Ratio) performance between pathology models. Compared to RAP measures, connectivity-weighted NAP measure of SC had Ratio = 1.90% less prediction error, whereas connectivity-weighted NAP measures of functional connectome (FC) had Ratio = 8.27% more prediction error. c Group averaged SC connectivity-weighted NAP scores. d Correlation coefficient between cognitive decline and SC connectivity-weighted NAP scores across cortical regions of interest (ROIs) (Spearman partial correlation, controlling for age, sex, education, baseline cognition, and hemispherical mean cortical thickness). e Correlation coefficient between cognitive decline and RAP scores across cortical ROIs. NAP scores have higher values in basal cortex regions (c). Individual variability in NAP scores in basal cortex regions and the left dorsolateral prefrontal cortex scales with cognitive decline (d), showing a stronger negative correlation with subsequent cognitive change compared to RAP scores (e).

Fig. 4:. The amyloid-β (Aβ) neuropathological signatures of future cognitive decline and external validation.

a Regions of interest (ROIs) whose regional Aβ neuropathology (RAP) measures were identified as features of future cognitive decline. This pattern of pathology signature was derived from the internal CogRes/RANN dataset. In external validation, features within the same ROIs pattern of ADNI participants were used to predict cognitive decline in delayed recall tests of memory (dMemory), and subsequently compared to the actual decline score. The predicted cognitive change significantly correlated with the actual score (R = 0.2583, p < 0.0285), and the relationship is significantly higher than the results obtained by random ROI selection (randomization p < 0.0140), demonstrating regional specificity. b Results of Aβ neuropathological signatures using connectivity-weighed network-based Aβ pathology (NAP) measures. c Results of Aβ neuropathological signatures using centrality-scaled NAP measures. The color indicates the network label of each ROI based on Schaefer atlas. Abbreviations: Cont, Cognitive control network; DorsAttn, Dorsal attention network; VisCent, Visual central network; VisPeri, Visual peripheral network; SalVentAttn, Salience/Ventral Attention network; Default, Default mode network.

Fig. 5:. Correlation results in the relationship between pathology features in each region of interest (ROI) and longitudinal cognitive decline.

a For participants in the cognitive reserve and reference ability neural network (CogRes/RANN) longitudinal study, we performed Spearman partial correlation analysis between pathology features in each ROI and longitudinal cognitive changes, controlling for age, sex, education, baseline cognition, and mean cortical thickness of each hemisphere. Pathology features were characterized by either regional amyloid-β pathology (RAP) or network-based amyloid-β pathology (NAP). b Same approach was performed for participants in the ADNI study. Compared to RAP measures, NAP scores demonstrated a significantly stronger negative relationship with cognition change in both studies. Cognition was assessed using global cognition in CogRes/RANN study and dMemory in ADNI study. (** p < 0.01; * p < 0.05; two-sample Student’s t-test)