Improving 3D convolutional neural network comprehensibility via interactive visualization of relevance maps: evaluation in Alzheimer's disease

ArcheD, a residual neural network for prediction of cerebrospinal fluid amyloid-beta from amyloid PET images

Detection and measurement of amyloid-beta (Aβ) in the brain is a key factor for early identification and diagnosis of Alzheimer's disease (AD). We aimed to develop a deep learning model to predict Aβ cerebrospinal fluid (CSF) concentration directly from amyloid PET images, independent of tracers, brain reference regions or preselected regions of interest. We used 1870 Aβ PET images and CSF measurements to train and validate a convolutional neural network ("ArcheD"). We evaluated the ArcheD performance in relation to episodic memory and the standardized uptake value ratio (SUVR) of cortical Aβ. We also compared the brain region's relevance for the model's CSF prediction within clinical-based and biological-based classifications. ArcheD-predicted Aβ CSF values correlated with measured Aβ CSF values (r = 0.92; q < 0.01), SUVR (rAV45 = -0.64, rFBB = -0.69; q < 0.01) and episodic memory measures (0.33 < r < 0.44; q < 0.01). For both classifications, cerebral white matter significantly contributed to CSF prediction (q < 0.01), specifically in non-symptomatic and early stages of AD. However, in late-stage disease, the brain stem, subcortical areas, cortical lobes, limbic lobe and basal forebrain made more significant contributions (q < 0.01). Considering cortical grey matter separately, the parietal lobe was the strongest predictor of CSF amyloid levels in those with prodromal or early AD, while the temporal lobe played a more crucial role for those with AD. In summary, ArcheD reliably predicted Aβ CSF concentration from Aβ PET scans, offering potential clinical utility for Aβ level determination.

Constructing personalized characterizations of structural brain aberrations in patients with dementia using explainable artificial intelligence

Deep learning approaches for clinical predictions based on magnetic resonance imaging data have shown great promise as a translational technology for diagnosis and prognosis in neurological disorders, but its clinical impact has been limited. This is partially attributed to the opaqueness of deep learning models, causing insufficient understanding of what underlies their decisions. To overcome this, we trained convolutional neural networks on structural brain scans to differentiate dementia patients from healthy controls, and applied layerwise relevance propagation to procure individual-level explanations of the model predictions. Through extensive validations we demonstrate that deviations recognized by the model corroborate existing knowledge of structural brain aberrations in dementia. By employing the explainable dementia classifier in a longitudinal dataset of patients with mild cognitive impairment, we show that the spatially rich explanations complement the model prediction when forecasting transition to dementia and help characterize the biological manifestation of disease in the individual brain. Overall, our work exemplifies the clinical potential of explainable artificial intelligence in precision medicine.

Distance-based novelty detection model for identifying individuals at risk of developing Alzheimer's disease

Introduction

Novelty detection (ND, also known as one-class classification) is a machine learning technique used to identify patterns that are typical of the majority class and can discriminate deviations as novelties. In the context of Alzheimer's disease (AD), ND could be employed to detect abnormal or atypical behavior that may indicate early signs of cognitive decline or the presence of the disease. To date, few research studies have used ND to discriminate the risk of developing AD and mild cognitive impairment (MCI) from healthy controls (HC).

Methods

In this work, two distinct cohorts with highly heterogeneous data, derived from the Australian Imaging Biomarkers and Lifestyle (AIBL) Flagship Study of Ageing project and the Fujian Medical University Union Hospital (FMUUH) China, were employed. An innovative framework with built-in easily interpretable ND models constructed solely on HC data was introduced along with proposing a strategy of distance to boundary (DtB) to detect MCI and AD. Subsequently, a web-based graphical user interface (GUI) that incorporates the proposed framework was developed for non-technical stakeholders.

Results

Our experimental results indicate that the best overall performance of detecting AD individuals in AIBL and FMUUH datasets was obtained by using the Mixture of Gaussian-based ND algorithm applied to single modality, with an AUC of 0.8757 and 0.9443, a sensitivity of 96.79% and 89.09%, and a specificity of 89.63% and 90.92%, respectively.

Discussion

The GUI offers an interactive platform to aid stakeholders in making diagnoses of MCI and AD, enabling streamlined decision-making processes. More importantly, the proposed DtB strategy could visually and quantitatively identify individuals at risk of developing AD.

Longitudinal trajectories of cognitive reserve in hypometabolic subtypes of Alzheimer's disease

Previous studies have demonstrated resilience to AD-related neuropathology in a form of cognitive reserve (CR). In this study we investigated a relationship between CR and hypometabolic subtypes of AD, specifically the typical and the limbic-predominant subtypes. We analyzed data from 59 Aβ-positive cognitively normal (CN), 221 prodromal Alzheimer's disease (AD) and 174 AD dementia participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) from ADNI and ADNIGO/2 phases. For replication, we analyzed data from 5 Aβ-positive CN, 89 prodromal AD and 43 AD dementia participants from ADNI3. CR was estimated as standardized residuals in a model predicting cognition from temporoparietal grey matter volumes and covariates. Higher CR estimates predicted slower cognitive decline. Typical and limbic-predominant hypometabolic subtypes demonstrated similar baseline CR, but the results suggested a faster decline of CR in the typical subtype. These findings support the relationship between subtypes and CR, specifically longitudinal trajectories of CR. Results also underline the importance of longitudinal analyses in research on CR.

Disentangling Accelerated Cognitive Decline from the Normal Aging Process and Unraveling Its Genetic Components: A Neuroimaging-Based Deep Learning Approach

Background

The progressive cognitive decline, an integral component of Alzheimer's disease (AD), unfolds in tandem with the natural aging process. Neuroimaging features have demonstrated the capacity to distinguish cognitive decline changes stemming from typical brain aging and AD between different chronological points.

Objective

To disentangle the normal aging effect from the AD-related accelerated cognitive decline and unravel its genetic components using a neuroimaging-based deep learning approach.

Methods

We developed a deep-learning framework based on a dual-loss Siamese ResNet network to extract fine-grained information from the longitudinal structural magnetic resonance imaging (MRI) data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study. We then conducted genome-wide association studies (GWAS) and post-GWAS analyses to reveal the genetic basis of AD-related accelerated cognitive decline.

Results

We used our model to process data from 1,313 individuals, training it on 414 cognitively normal people and predicting cognitive assessment for all participants. In our analysis of accelerated cognitive decline GWAS, we identified two genome-wide significant loci: APOE locus (chromosome 19 p13.32) and rs144614292 (chromosome 11 p15.1). Variant rs144614292 (G > T) has not been reported in previous AD GWA studies. It is within the intronic region of NELL1, which is expressed in neurons and plays a role in controlling cell growth and differentiation. The cell-type-specific enrichment analysis and functional enrichment of GWAS signals highlighted the microglia and immune-response pathways.

Conclusions

Our deep learning model effectively extracted relevant neuroimaging features and predicted individual cognitive decline. We reported a novel variant (rs144614292) within the NELL1 gene.

Predicting Four-Year's Alzheimer's Disease Onset Using Longitudinal Neurocognitive Tests and MRI Data Using Explainable Deep Convolutional Neural Networks

BACKGROUND

Prognosis of future risk of dementia from neuroimaging and cognitive data is important for optimizing clinical management for patients at early stage of Alzheimer's disease (AD). However, existing studies lack an efficient way to integrate longitudinal information from both modalities to improve prognosis performance.

OBJECTIVE

In this study, we aim to develop and evaluate an explainable deep learning-based framework to predict mild cognitive impairment (MCI) to AD conversion within four years using longitudinal whole-brain 3D MRI and neurocognitive tests.

METHODS

We proposed a two-stage framework that first uses a 3D convolutional neural network to extract single-timepoint MRI-based AD-related latent features, followed by multi-modal longitudinal feature concatenation and a 1D convolutional neural network to predict the risk of future dementia onset in four years.

RESULTS

The proposed deep learning framework showed promising to predict MCI to AD conversion within 4 years using longitudinal whole-brain 3D MRI and cognitive data without extracting regional brain volumes or cortical thickness, reaching a balanced accuracy of 0.834, significantly improved from models trained from single timepoint or single modality. The post hoc model explainability revealed heatmap indicating regions that are important for predicting future risk of AD.

CONCLUSIONS

The proposed framework sets the stage for future studies for using multi-modal longitudinal data to achieve optimal prediction for prognosis of AD onset, leading to better management of the diseases, thereby improving the quality of life.

Disentangling accelerated cognitive decline from the normal aging process and unraveling its genetic components: A neuroimaging-based deep learning approach

Background

The progressive cognitive decline that is an integral component of AD unfolds in tandem with the natural aging process. Neuroimaging features have demonstrated the capacity to distinguish cognitive decline changes stemming from typical brain aging and Alzheimer's disease between different chronological points.

Methods

We developed a deep-learning framework based on dual-loss Siamese ResNet network to extract fine-grained information from the longitudinal structural magnetic resonance imaging (MRI) data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) study. We then conducted genome-wide association studies (GWAS) and post-GWAS analyses to reveal the genetic basis of AD-related accelerated cognitive decline.

Results

We used our model to process data from 1,313 individuals, training it on 414 cognitively normal people and predicting cognitive assessment for all participants. In our analysis of accelerated cognitive decline GWAS, we identified two genome-wide significant loci: APOE locus (chromosome 19 p13.32) and rs144614292 (chromosome 11 p15.1). Variant rs144614292 (G>T) has not been reported in previous AD GWA studies. It is within the intronic region of NELL1, which is expressed in neuron and plays a role in controlling cell growth and differentiation. In addition, MUC7 and PROL1/OPRPNon chromosome 4 were significant at the gene level. The cell-type-specific enrichment analysis and functional enrichment of GWAS signals highlighted the microglia and immune-response pathways. Furthermore, we found that the cognitive decline slope GWAS was positively correlated with previous AD GWAS.

Conclusion

Our deep learning model was demonstrated effective on extracting relevant neuroimaging features and predicting individual cognitive decline. We reported a novel variant (rs144614292) within the NELL1 gene. Our approach has the potential to disentangle accelerated cognitive decline from the normal aging process and to determine its related genetic factors, leveraging opportunities for early intervention.

Distinct brain morphometry patterns revealed by deep learning improve prediction of aphasia severity

Emerging evidence suggests that post-stroke aphasia severity depends on the integrity of the brain beyond the stroke lesion. While measures of lesion anatomy and brain integrity combine synergistically to explain aphasic symptoms, significant interindividual variability remains unaccounted for. A possible explanatory factor may be the spatial distribution of brain atrophy beyond the lesion. This includes not just the specific brain areas showing atrophy, but also distinct three-dimensional patterns of atrophy. Here, we tested whether deep learning with Convolutional Neural Networks (CNN) on whole brain morphometry (i.e., segmented tissue volumes) and lesion anatomy can better predict which individuals with chronic stroke (N=231) have severe aphasia, and whether encoding spatial dependencies in the data might be capable of improving predictions by identifying unique individualized spatial patterns. We observed that CNN achieves significantly higher accuracy and F1 scores than Support Vector Machine (SVM), even when the SVM is nonlinear or integrates linear and nonlinear dimensionality reduction techniques. Performance parity was only achieved when the SVM was directly trained on the latent features learned by the CNN. Saliency maps demonstrated that the CNN leveraged widely distributed patterns of brain atrophy predictive of aphasia severity, whereas the SVM focused almost exclusively on the area around the lesion. Ensemble clustering of CNN saliency maps revealed distinct morphometry patterns that were unrelated to lesion size, highly consistent across individuals, and implicated unique brain networks associated with different cognitive processes as measured by the wider neuroimaging literature. Individualized predictions of severity depended on both ipsilateral and contralateral features outside of the location of stroke. Our findings illustrate that three-dimensional network distributions of atrophy in individuals with aphasia are directly associated with aphasia severity, underscoring the potential for deep learning to improve prognostication of behavioral outcomes from neuroimaging data, and highlighting the prospective benefits of interrogating spatial dependence at different scales in multivariate feature space.

Automated Medical Diagnosis of Alzheimer´s Disease Using an Efficient Net Convolutional Neural Network

Alzheimer's disease (AD) poses an enormous challenge to modern healthcare. Since 2017, researchers have been using deep learning (DL) models for the early detection of AD using neuroimaging biomarkers. In this paper, we implement the EfficietNet-b0 convolutional neural network (CNN) with a novel approach-"fusion of end-to-end and transfer learning"-to classify different stages of AD. 245 T1W MRI scans of cognitively normal (CN) subjects, 229 scans of AD subjects, and 229 scans of subjects with stable mild cognitive impairment (sMCI) were employed. Each scan was preprocessed using a standard pipeline. The proposed models were trained and evaluated using preprocessed scans. For the sMCI vs. AD classification task we obtained 95.29% accuracy and 95.35% area under the curve (AUC) for model training and 93.10% accuracy and 93.00% AUC for model testing. For the multiclass AD vs. CN vs. sMCI classification task we obtained 85.66% accuracy and 86% AUC for model training and 87.38% accuracy and 88.00% AUC for model testing. Based on our experimental results, we conclude that CNN-based DL models can be used to analyze complicated MRI scan features in clinical settings.

Visual deep learning of unprocessed neuroimaging characterises dementia subtypes and generalises across non-stereotypic samples

BACKGROUND

Dementia's diagnostic protocols are mostly based on standardised neuroimaging data collected in the Global North from homogeneous samples. In other non-stereotypical samples (participants with diverse admixture, genetics, demographics, MRI signals, or cultural origins), classifications of disease are difficult due to demographic and region-specific sample heterogeneities, lower quality scanners, and non-harmonised pipelines.

METHODS

We implemented a fully automatic computer-vision classifier using deep learning neural networks. A DenseNet was applied on raw (unpreprocessed) data from 3000 participants (behavioural variant frontotemporal dementia-bvFTD, Alzheimer's disease-AD, and healthy controls; both male and female as self-reported by participants). We tested our results in demographically matched and unmatched samples to discard possible biases and performed multiple out-of-sample validations.

FINDINGS

Robust classification results across all groups were achieved from standardised 3T neuroimaging data from the Global North, which also generalised to standardised 3T neuroimaging data from Latin America. Moreover, DenseNet also generalised to non-standardised, routine 1.5T clinical images from Latin America. These generalisations were robust in samples with heterogenous MRI recordings and were not confounded by demographics (i.e., were robust in both matched and unmatched samples, and when incorporating demographic variables in a multifeatured model). Model interpretability analysis using occlusion sensitivity evidenced core pathophysiological regions for each disease (mainly the hippocampus in AD, and the insula in bvFTD) demonstrating biological specificity and plausibility.

INTERPRETATION

The generalisable approach outlined here could be used in the future to aid clinician decision-making in diverse samples.

FUNDING

The specific funding of this article is provided in the acknowledgements section.

MPS-FFA: A multiplane and multiscale feature fusion attention network for Alzheimer's disease prediction with structural MRI

Structural magnetic resonance imaging (sMRI) is a popular technique that is widely applied in Alzheimer's disease (AD) diagnosis. However, only a few structural atrophy areas in sMRI scans are highly associated with AD. The degree of atrophy in patients' brain tissues and the distribution of lesion areas differ among patients. Therefore, a key challenge in sMRI-based AD diagnosis is identifying discriminating atrophy features. Hence, we propose a multiplane and multiscale feature-level fusion attention (MPS-FFA) model. The model has three components, (1) A feature encoder uses a multiscale feature extractor with hybrid attention layers to simultaneously capture and fuse multiple pathological features in the sagittal, coronal, and axial planes. (2) A global attention classifier combines clinical scores and two global attention layers to evaluate the feature impact scores and balance the relative contributions of different feature blocks. (3) A feature similarity discriminator minimizes the feature similarities among heterogeneous labels to enhance the ability of the network to discriminate atrophy features. The MPS-FFA model provides improved interpretability for identifying discriminating features using feature visualization. The experimental results on the baseline sMRI scans from two databases confirm the effectiveness (e.g., accuracy and generalizability) of our method in locating pathological locations. The source code is available at https://github.com/LiuFei-AHU/MPSFFA.

Modern Methods of Diagnostics and Treatment of Neurodegenerative Diseases and Depression

This paper discusses the promising areas of research into machine learning applications for the prevention and correction of neurodegenerative and depressive disorders. These two groups of disorders are among the leading causes of decline in the quality of life in the world when estimated using disability-adjusted years. Despite decades of research, the development of new approaches for the assessment (especially pre-clinical) and correction of neurodegenerative diseases and depressive disorders remains among the priority areas of research in neurophysiology, psychology, genetics, and interdisciplinary medicine. Contemporary machine learning technologies and medical data infrastructure create new research opportunities. However, reaching a consensus on the application of new machine learning methods and their integration with the existing standards of care and assessment is still a challenge to overcome before the innovations could be widely introduced to clinics. The research on the development of clinical predictions and classification algorithms contributes towards creating a unified approach to the use of growing clinical data. This unified approach should integrate the requirements of medical professionals, researchers, and governmental regulators. In the current paper, the current state of research into neurodegenerative and depressive disorders is presented.

A review of the application of three-dimensional convolutional neural networks for the diagnosis of Alzheimer's disease using neuroimaging

Alzheimer's disease (AD) is a degenerative disorder that leads to progressive, irreversible cognitive decline. To obtain an accurate and timely diagnosis and detect AD at an early stage, numerous approaches based on convolutional neural networks (CNNs) using neuroimaging data have been proposed. Because 3D CNNs can extract more spatial discrimination information than 2D CNNs, they have emerged as a promising research direction in the diagnosis of AD. The aim of this article is to present the current state of the art in the diagnosis of AD using 3D CNN models and neuroimaging modalities, focusing on the 3D CNN architectures and classification methods used, and to highlight potential future research topics. To give the reader a better overview of the content mentioned in this review, we briefly introduce the commonly used imaging datasets and the fundamentals of CNN architectures. Then we carefully analyzed the existing studies on AD diagnosis, which are divided into two levels according to their inputs: 3D subject-level CNNs and 3D patch-level CNNs, highlighting their contributions and significance in the field. In addition, this review discusses the key findings and challenges from the studies and highlights the lessons learned as a roadmap for future research. Finally, we summarize the paper by presenting some major findings, identifying open research challenges, and pointing out future research directions.

Can T1-Weighted Magnetic Resonance Imaging Significantly Improve Mini-Mental State Examination-Based Distinguishing Between Mild Cognitive Impairment and Early-Stage Alzheimer's Disease?

BACKGROUND

Detecting early-stage Alzheimer's disease (AD) is still problematic in clinical practice. This work aimed to find T1-weighted MRI-based markers for AD and mild cognitive impairment (MCI) to improve the screening process.

OBJECTIVE

Our assumption was to build a screening model that would be accessible and easy to use for physicians in their daily clinical routine.

METHODS

The multinomial logistic regression was used to detect status: AD, MCI, and normal control (NC) combined with the Bayesian information criterion for model selection. Several T1-weighted MRI-based radiomic features were considered explanatory variables in the prediction model.

RESULTS

The best radiomic predictor was the relative brain volume. The proposed method confirmed its quality by achieving a balanced accuracy of 95.18%, AUC of 93.25%, NPV of 97.93%, and PPV of 90.48% for classifying AD versus NC for the European DTI Study on Dementia (EDSD). The comparison of the two models: with the MMSE score only as an independent variable and corrected for the relative brain value and age, shows that the addition of the T1-weighted MRI-based biomarker improves the quality of MCI detection (AUC: 67.04% versus 71.08%) while maintaining quality for AD (AUC: 93.35% versus 93.25%). Additionally, among MCI patients predicted as AD inconsistently with the original diagnosis, 60% from ADNI and 76.47% from EDSD were re-diagnosed as AD within a 48-month follow-up. It shows that our model can detect AD patients a few years earlier than a standard medical diagnosis.

CONCLUSION

The created method is non-invasive, inexpensive, clinically accessible, and efficiently supports AD/MCI screening.

Structural and Functional MRI Data Differentially Predict Chronological Age and Behavioral Memory Performance

Human cognitive abilities decline with increasing chronological age, with decreased explicit memory performance being most strongly affected. However, some older adults show "successful aging," that is, relatively preserved cognitive ability in old age. One explanation for this could be higher brain-structural integrity in these individuals. Alternatively, the brain might recruit existing resources more efficiently or employ compensatory cognitive strategies. Here, we approached this question by testing multiple candidate variables from structural and functional neuroimaging for their ability to predict chronological age and memory performance, respectively. Prediction was performed using support vector machine (SVM) classification and regression across and within two samples of young (N = 106) and older (N = 153) adults. The candidate variables were (1) behavioral response frequencies in an episodic memory test; (2) recently described functional magnetic resonance imaging (fMRI) scores reflecting preservation of functional memory networks; (3) whole-brain fMRI contrasts for novelty processing and subsequent memory; (4) resting-state fMRI maps quantifying voxel-wise signal fluctuation; and (5) gray matter volume estimated from structural MRIs. While age group could be reliably decoded from all variables, chronological age within young and older subjects was best predicted from gray matter volume. In contrast, memory performance was best predicted from task-based fMRI contrasts and particularly single-value fMRI scores, whereas gray matter volume has no predictive power with respect to memory performance in healthy adults. Our results suggest that superior memory performance in healthy older adults is better explained by efficient recruitment of memory networks rather than by preserved brain structure.

Automated Evaluation of Conventional Clock-Drawing Test Using Deep Neural Network: Potential as a Mass Screening Tool to Detect Individuals With Cognitive Decline

Introduction

The Clock-Drawing Test (CDT) is a simple cognitive tool to examine multiple domains of cognition including executive function. We aimed to build a CDT-based deep neural network (DNN) model using data from a large cohort of older adults, to automatically detect cognitive decline, and explore its potential as a mass screening tool.

Methods

Over 40,000 CDT images were obtained from the National Health and Aging Trends Study (NHATS) database, which collects the annual surveys of nationally representative community-dwelling older adults in the United States. A convolutional neural network was utilized in deep learning architecture to predict the cognitive status of participants based on drawn clock images.

Results

The trained DNN model achieved balanced accuracy of 90.1 ± 0.6% in identifying those with a decline in executive function compared to those without [positive likelihood ratio (PLH) = 16.3 ± 6.8, negative likelihood ratio (NLH) = 0.14 ± 0.03], and 77.2 ± 2.7 % balanced accuracy for identifying those with probable dementia from those without (PLH = 5.1 ± 0.5, NLH = 0.37 ± 0.07).

Conclusions

This study demonstrated the feasibility of implementing conventional CDT to be automatically evaluated by DNN with a fair performance in a larger scale than ever, suggesting its potential as a mass screening test for ruling-in or ruling-out those with executive dysfunction or with probable dementia.

A systematic analysis of diagnostic performance for Alzheimer's disease using structural MRI

Background

Alzheimer's disease (AD) is one of the most common neurodegenerative disorders in the elderly. Although numerous structural magnetic resonance imaging (sMRI) studies have reported diagnostic models that could distinguish AD from normal controls (NCs) with 80-95% accuracy, limited efforts have been made regarding the clinically practical computer-aided diagnosis (CAD) system for AD.

Objective

To explore the potential factors that hinder the clinical translation of the AD-related diagnostic models based on sMRI.

Methods

To systematically review the diagnostic models for AD based on sMRI, we identified relevant studies published in the past 15 years on PubMed, Web of Science, Scopus, and Ovid. To evaluate the heterogeneity and publication bias among those studies, we performed subgroup analysis, meta-regression, Begg's test, and Egger's test.

Results

According to our screening criterion, 101 studies were included. Our results demonstrated that high diagnostic accuracy for distinguishing AD from NC was obtained in recently published studies, accompanied by significant heterogeneity. Meta-analysis showed that many factors contributed to the heterogeneity of high diagnostic accuracy of AD using sMRI, which included but was not limited to the following aspects: (i) different datasets; (ii) different machine learning models, e.g. traditional machine learning or deep learning model; (iii) different cross-validation methods, e.g. k-fold cross-validation leads to higher accuracies than leave-one-out cross-validation, but both overestimate the accuracy when compared to validation in independent samples; (iv) different sample sizes; and (v) the publication times. We speculate that these complicated variables might be the adverse factor for developing a clinically applicable system for the early diagnosis of AD.

Conclusions

Our findings proved that previous studies reported promising results for classifying AD from NC with different models using sMRI. However, considering the many factors hindering clinical radiology practice, there would still be a long way to go to improve.

Deep learning identifies brain structures that predict cognition and explain heterogeneity in cognitive aging

Specific brain structures (gray matter regions and white matter tracts) play a dominant role in determining cognitive decline and explain the heterogeneity in cognitive aging. Identification of these structures is crucial for screening of older adults at risk of cognitive decline. Using deep learning models augmented with a model-interpretation technique on data from 1432 Mayo Clinic Study of Aging participants, we identified a subset of brain structures that were most predictive of individualized cognitive trajectories and indicative of cognitively resilient vs. vulnerable individuals. Specifically, these structures explained why some participants were resilient to the deleterious effects of elevated brain amyloid and poor vascular health. Of these, medial temporal lobe and fornix, reflective of age and pathology-related degeneration, and corpus callosum, reflective of inter-hemispheric disconnection, accounted for 60% of the heterogeneity explained by the most predictive structures. Our results are valuable for identifying cognitively vulnerable individuals and for developing interventions for cognitive decline.