Background: Machine learning is a promising tool for biomarker-based diagnosis of Alzheimer’s disease (AD). Performing multimodal feature selection and studying the interaction between biological and clinical AD can help to improve the performance of the diagnosis models. Objective: This study aims to formulate a feature ranking metric based on the mutual information index to assess the relevance and redundancy of regional biomarkers and improve the AD classification accuracy. Methods: From the Alzheimer’s Disease Neuroimaging Initiative (ADNI), 722 participants with three modalities, including florbetapir-PET, flortaucipir-PET, and MRI, were studied. The multivariate mutual information metric was utilized to capture the redundancy and complementarity of the predictors and develop a feature ranking approach. This was followed by evaluating the capability of single-modal and multimodal biomarkers in predicting the cognitive stage. Results: Although amyloid-β deposition is an earlier event in the disease trajectory, tau PET with feature selection yielded a higher early-stage classification F1-score (65.4%) compared to amyloid-β PET (63.3%) and MRI (63.2%). The SVC multimodal scenario with feature selection improved the F1-score to 70.0% and 71.8% for the early and late-stage, respectively. When age and risk factors were included, the scores improved by 2 to 4%. The Amyloid-Tau-Neurodegeneration [AT(N)] framework helped to interpret the classification results for different biomarker categories. Conclusion: The results underscore the utility of a novel feature selection approach to reduce the dimensionality of multimodal datasets and enhance model performance. The AT(N) biomarker framework can help to explore the misclassified cases by revealing the relationship between neuropathological biomarkers and cognition.
more »
« less
What’s the cut-point?: a systematic investigation of tau PET thresholding methods
Tau positron emission tomography (PET) is increasing in popularity for biomarker characterization of Alzheimer’s disease (AD), and recent frameworks rely on tau PET cut-points to stage individuals along the AD continuum. Given the lack of standardization in tau PET thresholding methods, this study sought to systematically canvass and characterize existing studies that have derived tau PET cut-points and then directly assess different methods of tau PET thresholding in terms of their concurrent validity.
First, a literature search was conducted in PubMed to identify studies of AD and related clinical phenotypes that used the Flortaucipir (AV-1451) tau PET tracer to derive a binary cut-point for tau positivity. Of 540 articles screened and 47 full-texts reviewed, 23 cohort studies met inclusion criteria with a total of 6536 participants. Second, we derived and compared tau PET cut-points in a 2 × 2 × 2 design that systematically varied region (temporal meta-ROI and entorhinal cortex), analytic method (receiver operating characteristics and 2 standard deviations above comparison group), and criterion/comparison variable (amyloid-beta negative cognitively unimpaired or cognitively unimpaired only) using a sample of 453 older adults from the Alzheimer’s Disease Neuroimaging Initiative.
For the systematic review, notable variability in sample characteristics, preprocessing methods, region of interest, and analytic approach were observed, which were accompanied by discrepancy in proposed tau PET cut points. The empirical follow-up indicated the cut-point derived based on 2 standard deviations above a either comparison group in either ROI best differentiated tau positive and negative groups on cerebrospinal fluid phosphorylated tau, Mini-Mental State Examination score, and delayed memory performance.
Given the impact of discrepant thresholds on tau positivity rates, biomarker staging, and eligibility for future clinical treatment trials, recommendations are offered to select cut-point derivations based on the unique goals and priorities of different studies.
- NSF-PAR ID:
- 10365379
- Publisher / Repository:
- Springer Science + Business Media
- Date Published:
- Journal Name:
- Alzheimer's Research & Therapy
- Volume:
- 14
- Issue:
- 1
- ISSN:
- 1758-9193
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
more » « less
Abstract INTRODUCTION A biphasic model for brain structural changes in preclinical Alzheimer's disease (AD) could reconcile some conflicting and paradoxical findings in observational studies and anti‐amyloid clinical trials.
METHODS In this study we tested this model fitting linear versus quadratic trajectories and computed the timing of the inflection points vertexwise of cortical thickness and cortical diffusivity—a novel marker of cortical microstructure—changes in 389 participants from the Dominantly Inherited Alzheimer Network.
RESULTS In early preclinical AD, between 20 and 15 years before estimated symptom onset, we found increases in cortical thickness and decreases in cortical diffusivity followed by cortical thinning and cortical diffusivity increases in later preclinical and symptomatic stages. The inflection points 16 to 19 years before estimated symptom onset are in agreement with the start of tau biomarker alterations.
DISCUSSION These findings confirm a biphasic trajectory for brain structural changes and have direct implications when interpreting magnetic resonance imaging measures in preventive AD clinical trials.
-
null (Ed.)Abstract The ability to measure total and phosphorylated tau levels in clinical samples is transforming the detection of Alzheimer’s disease (AD) and other neurodegenerative diseases. In particular, recent reports indicate that accurate detection of low levels of phosphorylated tau (p-tau) in plasma provides a reliable biomarker of AD long before sensing memory loss. Therefore, the diagnosis and monitoring of neurodegenerative diseases progression using blood samples is becoming a reality. These major advances were achieved by using antibodies specific to p-tau as well as sophisticated high-sensitivity immunoassay platforms. This review focuses on these enabling advances in high-specificity antibody development, engineering, and novel signal detection methods. We will draw insights from structural studies on p-tau antibodies, engineering efforts to improve their binding properties, and efforts to validate their specificity. A comprehensive survey of high-sensitivity p-tau immunoassay platforms along with sensitivity limits will be provided. We conclude that although robust approaches for detecting certain p-tau species have been established, systematic efforts to validate antibodies for assay development is still needed for the recognition of biomarkers for AD and other neurodegenerative diseases.more » « less
-
Abstract Neuroimaging in the preclinical phase of Alzheimer’s disease provides information crucial to early intervention, particularly in people with a high genetic risk. Metabolic network modularity, recently applied to the study of dementia, is increased in Alzheimer’s disease patients compared with controls, but network modularity in cognitively unimpaired elderly with various risks of developing Alzheimer’s disease needs to be determined. Based on their 5-year cognitive progression, we stratified 117 cognitively normal participants (78.3 ± 4.0 years of age, 52 women) into three age-matched groups, each with a different level of risk for Alzheimer’s disease. From their fluorodeoxyglucose PET we constructed metabolic networks, evaluated their modular structures using the Louvain algorithm, and compared them between risk groups. As the risk for Alzheimer’s disease increased, the metabolic connections among brain regions weakened and became more modular, indicating network fragmentation and functional impairment of the brain. We then set out to determine the correlation between regional brain metabolism, particularly in the modules derived from the previous analysis, and the regional expression of Alzheimer-risk genes in the brain, obtained from the Allen Human Brain Atlas. In all risk groups of this elderly population, the regional brain expression of most Alzheimer-risk genes showed a strong correlation with brain metabolism, particularly in the module that corresponded to regions of the brain that are affected earliest and most severely in Alzheimer’s disease. Among the genes, APOE and CD33 showed the strongest negative correlation and SORL1 showed the strongest positive correlation with brain metabolism. The Pearson correlation coefficients remained significant when contrasted against a null-hypothesis distribution of correlation coefficients across the whole transcriptome of 20 736 genes (SORL1: P = 0.0130; CD33, P = 0.0136; APOE: P = 0.0093). The strong regional correlation between Alzheimer-related gene expression in the brain and brain metabolism in older adults highlights the role of brain metabolism in the genesis of dementia.more » « less
-
more » « less
Abstract The dentate gyrus (DG), a key hippocampal subregion in memory processing, generally resists phosphorylated tau accumulation in the amnestic dementia of the Alzheimer’s type due to Alzheimer’s disease (DAT-AD), but less is known about the susceptibility of the DG to other tauopathies. Here, we report stereologic densities of total DG neurons and tau inclusions in thirty-two brains of human participants with autopsy-confirmed tauopathies with distinct isoform profiles—3R Pick’s disease (PiD, N = 8), 4R corticobasal degeneration (CBD, N = 8), 4R progressive supranuclear palsy (PSP, N = 8), and 3/4R AD (N = 8). All participants were diagnosed during life with primary progressive aphasia (PPA), an aphasic clinical dementia syndrome characterized by progressive deterioration of language abilities with spared non-language cognitive abilities in early stages, except for five patients with DAT-AD as a comparison group. 51% of total participants were female. All specimens were stained immunohistochemically with AT8 to visualize tau pathology, and PPA cases were stained for Nissl substance to visualize neurons. Unbiased stereological analysis was performed in granule and hilar DG cells, and inclusion-to-neuron ratios were calculated. In the PPA group, PiD had highest mean total (granule + hilar) densities of DG tau pathology (
p < 0.001), followed by CBD, AD, then PSP. PPA-AD cases showed more inclusions in hilar cells compared to granule cells, while the opposite was true in PiD and CBD. Inclusion-to-neuron ratios revealed, on average, 33% of all DG neurons in PiD cases contained a tau inclusion, compared to ~ 7% in CBD, 2% in AD, and 0.4% in PSP. There was no significant difference between DAT-AD and PPA-AD pathologic tau burden, suggesting that differences in DG burden are not specific to clinical phenotype. We conclude that the DG is differentially vulnerable to pathologic tau accumulation, raising intriguing questions about the structural integrity and functional significance of hippocampal circuits in neurodegenerative dementias.
