About

Ani Eloyan is an Associate Professor of Biostatistics and serves as Vice Chair of Biostatistics at Brown University. Her research interests include semi-parametric likelihood based methods for matrix decompositions, statistical analyses of brain images, and the integration of various types of complex data structures for analyzing healthcare data. She has contributed to the development of statistical methodologies for medical imaging, brain lesion modeling, and longitudinal biomarker detection, with applications spanning multiple sclerosis, Alzheimer's disease, epilepsy, autism, and other neurological conditions. Dr. Eloyan's work emphasizes the advancement of statistical techniques to better understand and interpret complex biomedical data, supporting improved diagnosis, disease progression analysis, and health outcomes.

Research topics

• Internal medicine
• Medicine
• Computer Science
• Artificial Intelligence
• Mathematics
• Statistics
• Psychology
• Neuroscience

Selected publications

• Distinct Tau PET Dynamics in Early vs. Late Age-of-Onset Alzheimer’s disease

  medRxiv · 2026-01-26

  articleOpen access

  Early-onset Alzheimer's disease (EOAD) and Late-onset AD (LOAD) differ in clinical presentations and rates of progression. We aimed to compare baseline and longitudinal tau PET burden, and their relationship with clinical variables in amyloid-PET positive, cognitively impaired participants from the Longitudinal Early-Onset Alzheimer's Disease Study (EOAD; n=390) and Alzheimer's Disease Neuroimaging Initiative (LOAD; n=211). Patients with EOAD showed higher baseline tau PET retention, broader neuroanatomical involvement and faster accumulation rates over time compared to LOAD, after adjusting for amyloid load and clinical stage. Tau PET showed stronger correlations with baseline amyloid burden and clinical measures of global cognition and function in EOAD than LOAD. We conclude that earlier age of onset in AD is linked to a more aggressive tauopathy, which in turn is a primary driver of clinical decline. These findings suggest that optimal therapeutic targets and strategies may differ between EOAD and LOAD.

  PublisherOA PDFDOI

• Evaluation of cortical pathology in primary-progressive multiple sclerosis: a post hoc analysis of the SPRINT-MS trial

  BMJ Neurology Open · 2026-01-01

  articleOpen access

  Background Multiple sclerosis (MS) involves cortical injury, including cortical lesion (CL) development. Ibudilast treatment was found to slow progression of whole brain and cortical atrophy, but the effect of ibudilast on CLs is unknown. The present study aims to evaluate the treatment effect of ibudilast on CL development and whether the effect of ibudilast on brain atrophy is modified by CLs in primary-progressive MS (PPMS). Methods In this longitudinal study, we analysed data of 102 people with PPMS (ibudilast: n=49; placebo: n=53) from the Secondary and Primary Progressive Ibudilast NeuroNEXT Trial in Multiple Sclerosis. CLs were identified on artificial intelligence-generated double inversion-recovery images created from T1 and T2 images at 3T MRI and rated at baseline and week 96. Atrophy was measured by cortical thickness and brain parenchymal fraction. Results CLs were detected in all participants with PPMS with a median count of 22 (11–34) in the ibudilast group and 28 (13–44) in the placebo group. At baseline, treatment groups did not differ in any brain volume measure or CL count. Higher CL counts at baseline were associated with higher CL formation during follow-up ( B(95% CI )=0.20 (0.12 to 0.29), p<0.001), independent of ibudilast treatment. Change in CL count did not differ between treatment groups ( mean difference ( SD )=0.07 (0.29), p=0.364). Protective effect of ibudilast on cortical thickness was more prominent in subjects with greater CL formation, but this relationship was not observed with whole brain atrophy. Conclusions Ibudilast treatment does not affect CL development in PPMS. Its protective effect on cortical thinning is more prominent with greater CL formation. Trial registration number NCT01982942 .

  PublisherOA PDFDOI

• TRAECR: A Tool for Preprocessing Positron Emission Tomography Imaging for Statistical Modeling

  ArXiv.org · 2025-11-06

  preprintOpen accessSenior author

  Positron emission tomography (PET) imaging is widely used in a number of clinical applications, including cancer and Alzheimer's disease (AD) diagnosis, monitoring of disease development, and treatment effect evaluation. Statistical modeling of PET imaging is essential to address continually emerging scientific questions in these research fields, including hypotheses related to evaluation of effects of disease modifying treatments on amyloid reduction in AD and associations between amyloid reduction and cognitive function, among many others. In this paper, we provide background information and tools for statisticians interested in developing statistical models for PET imaging to pre-process and prepare data for analysis. We introduce our novel pre-processing and visualization tool TRAECR (Template registration, MRI-PET co-Registration, Anatomical brain Extraction and COMBAT/RAVEL harmonization) to facilitate data preparation for statistical analysis.

  PublisherOA PDFDOI

• Voiceprints of cognitive impairment: analyzing digital voice for early detection of Alzheimer’s and related dementias

  npj Dementia · 2025-11-03 · 2 citations

  articleOpen access

  Early detection of Alzheimer's disease (AD) is critical yet challenging, particularly in younger individuals. This study leverages artificial intelligence to analyze digital voice recordings from the Craft Story Recall task within the Longitudinal Early-onset AD Study (LEADS) to (1) detect cognitive impairment and (2) differentiate early-onset AD (EOAD) from early onset non-AD cognitive impairment (EOnonAD). Using speech samples from 120 patients and 68 cognitively unimpaired controls, we employed two classification approaches: feature-engineered machine learning and end-to-end deep learning incorporating a Large Language Model. To detect mild cognitive impairment, the feature-engineered model, using acoustic and linguistic features, achieved an AUC of 0.945 on the holdout test set, while the end-to-end model yielded an AUC of 0.988. For differentiating EOAD from EOnonAD, the feature-engineered model achieved an AUC of 0.804, and the end-to-end model yielded an AUC of 0.904 on the holdout set. Explainability analyses revealed reduced linguistic informativeness as a key AD indicator.

  PublisherOA PDFDOI

• Geometric Modeling of Hippocampal Tau Deposition: A Surface-Based Framework for Covariate Analysis and Off-Target Contamination Detection

  ArXiv.org · 2025-11-03

  preprintOpen accessSenior author

  We introduce a framework combining geometric modeling with disease progression analysis to investigate tau deposition in Alzheimer's disease (AD) using positron emission tomography (PET) data. Focusing on the hippocampus, we construct a principal surface that captures the spatial distribution and morphological changes of tau pathology. By projecting voxels onto this surface, we quantify tau coverage, intensity, and thickness through bidirectional projection distances and interpolated standardized uptake value ratios (SUVR). This low-dimensional embedding preserves spatial specificity while mitigating multiple comparison issues. Covariate effects are analyzed using a two-stage regression model with inverse probability weighting to adjust for signal sparsity and selection bias. Using the SuStaIn model, we identify subtypes and stages of AD, revealing distinct tau dynamics: the limbic-predominant subtype shows age-related nonlinear accumulation in coverage and thickness, whereas the posterior subtype exhibits uniform SUVR increases across disease progression. Model-based predictions show that hippocampal tau deposition follows a structured spatial trajectory expanding bidirectionally with increasing thickness, while subtype differences highlight posterior hippocampal involvement consistent with whole-brain patterns. Finally, directional signal patterns on the principal surface reveal contamination from the choroid plexus, demonstrating the broader applicability of the proposed framework across modalities including amyloid PET.

  PublisherOA PDFDOI

• Characterizing and validating 12-month reliable cognitive change in Early-Onset Alzheimer's Disease for use in clinical trials

  The Journal of Prevention of Alzheimer s Disease · 2025-02-16 · 5 citations

  articleOpen access

  BACKGROUND: As literature suggests that Early-Onset Alzheimer's Disease (EOAD) and late-onset AD may differ in important ways, need exists for randomized clinical trials for treatments tailored to EOAD. Accurately measuring reliable cognitive change in individual patients with EOAD will have great value for these trials. OBJECTIVES: The current study sought to characterize and validate 12-month reliable change from the Longitudinal Early-Onset Alzheimer's Disease Study (LEADS) neuropsychological battery. DESIGN: Standardized regression-based (SRB) prediction equations were developed from age-matched cognitively intact participants within LEADS, and applied to clinically impaired participants from LEADS. SETTING: Participants were recruited from outpatient academic medical centers. PARTICIPANTS: Participants were enrolled in LEADS and diagnosed with amyloid-positive EOAD (n = 189) and amyloid-negative early-onset cognitive impairment not related to AD (EOnonAD; n = 43). MEASUREMENT: 12-month reliable change (Z-scores) was compared between groups across cognitive domain composites, and distributions of individual participant trajectories were examined. Prediction of Z-scores by common AD biomarkers was also considered. RESULTS: Both EOAD and EOnonAD displayed significantly lower 12-month follow-up scores than were predicted based on SRB equations, with declines more pronounced for EOAD across several domains. AD biomarkers of cerebral β-amyloid, tau, and EOAD-specific atrophy were predictive of 12-month change scores. CONCLUSIONS: The current results support including EOAD patients in longitudinal clinical trials, and generate evidence of validation for using 12-month reliable cognitive change as a clinical outcome metric in clinical trials in EOAD cohorts like LEADS. Doing so will enhance the success of EOAD trials and permit a better understanding of individual responses to treatment.

  PublisherDOI

• EOAD‐Signature Atrophy Predicts Progression to Dementia in Patients with Early‐onset MCI due to Alzheimer's Disease

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  Abstract Background Prognostic risk stratification for patients at the mild cognitive impairment (MCI) stage of early‐onset Alzheimer's disease (EOAD) would allow professionals and loved ones to make better‐informed medical and life planning decisions. While research including our own (Bakkour, Morris, Dickerson, 2009) has demonstrated the prognostic value of MRI‐based measures of brain structure in late‐onset amnestic AD, its utility for predicting progression to dementia in EOAD remains unclear. Here, we measured the magnitude of cortical atrophy within our recently described EOAD signature regions (Touroutoglou et al. 2023) in patients with EOAD at the MCI stage ( N = 130) recruited in LEADS. The main goal of the study was to evaluate the utility of this measure as a predictor of time to subsequent progression to dementia. Our second goal was to examine the independent or synergistic contributions of EOAD signature of atrophy and standard clinical severity measures used in clinical trials. Method For each patient, we measured the time between baseline visit and subsequent visit at which progression to mild dementia was documented or last observation. Baseline cortical atrophy was measured as W ‐scores (i.e., Z ‐scores adjusted for age and sex relative to a sample of healthy controls) in the EOAD signature. Baseline clinical severity was quantified with the Clinical Dementia Rating Sum‐of‐Boxes scores (CDR‐SB). Simple and multivariable Cox regression models examined the relationship between atrophy in EOAD signature, baseline CDR‐SB, and the likelihood of progression to dementia. Result Greater baseline atrophy in the EOAD signature predicted higher risk of progression to dementia (hazard ratio = 1.2, 95% CI 1.1‐1.3) and provided additive value to the CDR‐SB (hazard ratio: 2.1, 95% CI:1.7‐2.8) in predicting progression. Conclusion These findings point to the role of EOAD MRI signature as an imaging biomarker to guide prognostication for patients with EOAD and their families and to inform the design of clinical trials.

  PublisherOA PDFDOI

• Dissociable spatial topography of cortical atrophy in early‐onset and late‐onset Alzheimer's disease: A head‐to‐head comparison of the LEADS and ADNI cohorts

  Alzheimer s & Dementia · 2025-02-01 · 8 citations

  articleOpen access

  INTRODUCTION: Early-onset and late-onset Alzheimer's disease (EOAD and LOAD, respectively) have distinct clinical manifestations, with prior work based on small samples suggesting unique patterns of neurodegeneration. The current study performed a head-to-head comparison of cortical atrophy in EOAD and LOAD, using two large and well-characterized cohorts (LEADS and ADNI). METHODS: We analyzed brain structural magnetic resonance imaging (MRI) data acquired from 377 sporadic EOAD patients and 317 sporadicLOAD patients who were amyloid positive and had mild cognitive impairment (MCI) or mild dementia (i.e., early-stage AD), along with cognitively unimpaired participants. RESULTS: After controlling for the level of cognitive impairment, we found a double dissociation between AD clinical phenotype and localization/magnitude of atrophy, characterized by predominant neocortical involvement in EOAD and more focal anterior medial temporal involvement in LOAD. DISCUSSION: Our findings point to the clinical utility of MRI-based biomarkers of atrophy in differentiating between EOAD and LOAD, which may be useful for diagnosis, prognostication, and treatment. HIGHLIGHTS: Early-onset Alzheimer's disease (EOAD) and late-onset AD (LOAD) patients showed distinct and overlapping cortical atrophy patterns. EOAD patients showed prominent atrophy in widespread neocortical regions. LOAD patients showed prominent atrophy in the anterior medial temporal lobe. Regional atrophy was correlated with the severity of global cognitive impairment. Results were comparable when the sample was stratified for mild cognitive impairment (MCI) and dementia.

  PublisherOA PDFDOI

• Clinical and cognitive manifestations of hemispheric asymmetry in neuroimaging biomarkers in early onset AD

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Asymmetrical imaging findings have been observed in individuals across the AD continuum. However, the prevalence and consequences of asymmetry in clinically heterogeneous early-onset AD (EOAD) is not well understood. METHOD: We included 373 amyloid-PET-positive patients under 65 years with MCI or mild dementia from the Longitudinal Early Onset Alzheimer's Disease Study. Patient clinical phenotypes were amnestic or nonamnestic EOAD, primary progressive aphasia (PPA), or posterior cortical atrophy (PCA). Participants had florbetaben and flortaucipir PET, MRI, and neuropsychological testing at baseline. For each imaging modality, we used Freesurfer7.1 segmentation to compute a global asymmetry index: left/right hemisphere signal difference in all cortical regions divided by bilateral signal. To examine spatial patterns of asymmetry, we used a symmetrical, sample-specific template to calculate an analogous asymmetry value in each PET/MRI voxel in a single hemisphere of the brain. RESULT: Global asymmetry was observed across modalities, most prominently in flortaucipir (Figure 1). Both left and right asymmetry was evident in all clinical phenotypes, though flortaucipir signal was markedly left-lateralized for PPA right-lateralized for PCA. Asymmetry was correlated between modalities, particularly flortaucipir and atrophy. Voxelwise analyses showed heterogeneous patterns of flortaucipir asymmetry in amnestic and nonamnestic groups (Figure 2). Left-lateralized signal was prominent across cortex except in tempoparietal regions in PPA, and right-lateralized signal was prominent except in occipital regions in PCA. For all clinical phenotypes, asymmetry was minimal in areas of highest flortaucipir signal (averaged across hemispheres): one-way ANOVA revealed nonoverlapping regions of group differences in asymmetry and average signal. Adjusting for age, sex, education, and overall cortical abnormality in the whole sample, flortaucipir asymmetry was most robustly associated with worse cognitive performance, followed by atrophy and florbetaben (Figure 3). Worse verbal fluency and working memory related to left-lateralized flortaucipir signal in frontal and occipital cortex. Worse performance in visuospatial tasks was associated with right-lateralized signal in frontotemporal areas; right-predominant medial temporal asymmetry correlated with worse delayed figure recall performance. CONCLUSION: Asymmetric neuroimaging patterns are common in EOAD and correlate with clinical presentation. Specific early cognitive deficits may be predicted by examining patterns of asymmetry with heterogeneous clinical and cognitive manifestations.

  PublisherOA PDFDOI

• Aberrant intrinsic functional network connectivity in Early‐Onset Alzheimer’s Disease

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Neurodegeneration in sporadic early-onset Alzheimer's disease (EOAD) at the stage of MCI and mild dementia is characterized by atrophy most prominently in posterior temporoparietal cortical areas. These "EOAD signature" regions (Touroutoglou et al., 2023) spatially correspond to key nodes of several canonical large-scale functional networks, including the default mode, frontoparietal, language, and dorsal attention networks. While current evidence points to abnormal functional connectivity in EOAD compared with healthy controls, prior studies employed small samples and have yielded mixed results. Here, we analyzed a large sample of sporadic EOAD patients from the Longitudinal Early-Onset Alzheimer's Disease Study (LEADS) to test the central hypothesis that EOAD disrupts the functional integrity of several brain networks. We also hypothesized that AD-related connectivity alteration would be associated with the magnitude of cognitive impairment and of cortical atrophy in the EOAD signature. METHOD: We analyzed whole-brain multi-band functional MRI data (duration = 10 min) collected at wakeful rest from sporadic EOAD (n = 239) and cognitively unimpaired (CU) (n = 68) participants from the LEADS cohort (Table 1). We used a custom pipeline to conduct preprocessing, confound removal, and quality control of MRI data. The residual functional timeseries data were used to calculate seed-based functional connectivity of each anatomically distinct cortical region within the EOAD signature (Figure 1A). We then compared these seed-based connectivity maps between EOAD and CU participants, while controlling for age, sex, and estimates of in-scanner head motion. RESULT: EOAD patients showed abnormally stronger connectivity in all EOAD signature regions than CU participants, involving multiple large-scale functional networks (Figure 1B). In EOAD patients, stronger functional connectivity within multiple networks was associated with worse cognitive impairment and greater atrophy in the EOAD signature (Figure 2). An exploratory analysis of whole-cortex functional connectome further showed that AD-related hyperconnectivity was prominently observed in relation to the default mode and frontoparietal networks. CONCLUSION: In patients with sporadic EOAD, phenotypically vulnerable cortical regions exhibit prominent disruption of intrinsic functional connectivity. Our findings support the selective vulnerability of cortical functional networks in EOAD, which may underlie the distinct pattern of AD pathology spreading and cognitive impairment in this AD clinical phenotype.

  PublisherOA PDFDOI

Recent grants

• Spatiotemporal Modeling of MRI Brain Lesion Trajectories of Biomarker Discovery

  NIH · $454k · 2016–2018

• Administrative Core

  NIH · $46.6M · 2013–2025

Frequent coauthors

• Liana G. Apostolova

  140 shared

• Joseph C. Masdeu

  Houston Methodist

  118 shared

• Gil D. Rabinovici

  University Memory and Aging Center

  106 shared

• Dustin B. Hammers

  Indiana University – Purdue University Indianapolis

  82 shared

• Tatiana Foroud

  Indiana University – Purdue University Indianapolis

  80 shared

• Brad C. Dickerson

  78 shared

• Chiadi U. Onyike

  Johns Hopkins University

  76 shared

• Kelly Nudelman

  74 shared