Faculty Bibliography

Early identification of Alzheimer's disease (AD) and AD-related dementias (ADRD) has high clinical significance, both because of the potential to slow decline through initiating FDA-approved therapies and managing modifiable risk factors, and to help persons living with dementia and their families to plan before cognitive loss makes doing so challenging. However, substantial racial and ethnic disparities in early diagnosis currently lead to additional inequities in care, urging accurate and inclusive risk assessment programs. In this study, we trained an artificial intelligence foundation model to represent the electronic health records (EHR) data with a vast cohort of 1.2 million patients within a large health system. Building upon this foundation EHR model, we developed a predictive Transformer model, named TRADE, capable of identifying risks for AD/ADRD and mild cognitive impairment (MCI), by analyzing the past sequential visit records. Amongst individuals 65 and older, our model was able to generate risk predictions for various future timeframes. On the held-out validation set, our model achieved an area under the receiver operating characteristic (AUROC) of 0.772 (95% CI: 0.770, 0.773) for identifying the AD/ADRD/MCI risks in 1 year, and AUROC of 0.735 (95% CI: 0.734, 0.736) in 5 years. The positive predictive values (PPV) in 5 years among individuals with top 1% and 5% highest estimated risks were 39.2% and 27.8%, respectively. These results demonstrate significant improvements upon the current EHR-based AD/ADRD/MCI risk assessment models, paving the way for better prognosis and management of AD/ADRD/MCI at scale.

BACKGROUND:We examined drivers of self and study partner reports of memory loss in mild cognitive impairment (MCI) from Alzheimer (AD-MCI) and vascular disease (Va-MCI). METHODS:We performed retrospective cross-sectional analyses of participants with AD-MCI (n=2874) and Va-MCI (n=376) from the National Alzheimer's Coordinating Center data set. Statistical analysis utilized 2-sided t test or the Fisher exact test. RESULTS:Compared with AD-MCI, Va-MCI subjects (24.5% vs. 19.7%, P=0.031) and study partners (31.4% vs. 21.6%, P<0.0001) were more likely to deny memory loss. Black/African Americans were disproportionately represented in the group denying memory loss in AD-MCI (20.0% vs. 13.2%, P<0.0001) and Va-MCI (33.7% vs. 18.0%, P=0.0022). Study partners of participants with these features also disproportionately denied memory loss: female (AD-MCI: 60.1% vs. 51.7%, P=0.0002; Va-MCI: 70.3% vs. 52.3%, P=0.0011), Black/African American (AD-MCI: 23.5% vs. 11.98%, P<0.0001; Va-MCI: 48.8% vs. 26.5%, P=0.0002), and <16 years of education (AD-MCI only: 33.9% vs. 16.3%, P=0.0262). In AD-MCI and Va-MCI, participants with anxiety were disproportionately represented in the group endorsing memory loss (AD: 28.2% vs. 17.4%, P<0.0001; Va: 31.5% vs. 16.1%, P=0.0071), with analogous results with depression. CONCLUSION/CONCLUSIONS:The findings would suggest extra vigilance in interview-based MCI detection of persons at-risk for self-based or informant-based misreport.

IMPORTANCE/UNASSIGNED:Large language models (LLMs) are crucial for medical tasks. Ensuring their reliability is vital to avoid false results. Our study assesses two state-of-the-art LLMs (ChatGPT and LlaMA-2) for extracting clinical information, focusing on cognitive tests like MMSE and CDR. OBJECTIVE/UNASSIGNED:Evaluate ChatGPT and LlaMA-2 performance in extracting MMSE and CDR scores, including their associated dates. METHODS/UNASSIGNED:Our data consisted of 135,307 clinical notes (Jan 12th, 2010 to May 24th, 2023) mentioning MMSE, CDR, or MoCA. After applying inclusion criteria 34,465 notes remained, of which 765 underwent ChatGPT (GPT-4) and LlaMA-2, and 22 experts reviewed the responses. ChatGPT successfully extracted MMSE and CDR instances with dates from 742 notes. We used 20 notes for fine-tuning and training the reviewers. The remaining 722 were assigned to reviewers, with 309 each assigned to two reviewers simultaneously. Inter-rater-agreement (Fleiss' Kappa), precision, recall, true/false negative rates, and accuracy were calculated. Our study follows TRIPOD reporting guidelines for model validation. RESULTS/UNASSIGNED:For MMSE information extraction, ChatGPT (vs. LlaMA-2) achieved accuracy of 83% (vs. 66.4%), sensitivity of 89.7% (vs. 69.9%), true-negative rates of 96% (vs 60.0%), and precision of 82.7% (vs 62.2%). For CDR the results were lower overall, with accuracy of 87.1% (vs. 74.5%), sensitivity of 84.3% (vs. 39.7%), true-negative rates of 99.8% (98.4%), and precision of 48.3% (vs. 16.1%). We qualitatively evaluated the MMSE errors of ChatGPT and LlaMA-2 on double-reviewed notes. LlaMA-2 errors included 27 cases of total hallucination, 19 cases of reporting other scores instead of MMSE, 25 missed scores, and 23 cases of reporting only the wrong date. In comparison, ChatGPT's errors included only 3 cases of total hallucination, 17 cases of wrong test reported instead of MMSE, and 19 cases of reporting a wrong date. CONCLUSIONS/UNASSIGNED:In this diagnostic/prognostic study of ChatGPT and LlaMA-2 for extracting cognitive exam dates and scores from clinical notes, ChatGPT exhibited high accuracy, with better performance compared to LlaMA-2. The use of LLMs could benefit dementia research and clinical care, by identifying eligible patients for treatments initialization or clinical trial enrollments. Rigorous evaluation of LLMs is crucial to understanding their capabilities and limitations.

INTRODUCTION/UNASSIGNED:Alzheimer's Disease (AD) typically starts in the medial temporal lobe, then develops into a neurodegenerative cascade which spreads to other brain regions. People with subjective cognitive decline (SCD) are more likely to develop dementia, especially in the presence of amyloid pathology. Thus, we were interested in the white matter microstructure of the medial temporal lobe in SCD, specifically the lower cingulum bundle that leads into the hippocampus. Diffusion tensor imaging (DTI) has been shown to differentiate SCD participants who will progress to mild cognitive impairment from those who will not. However, the biology underlying these DTI metrics is unclear, and results in the medial temporal lobe have been inconsistent. METHODS/UNASSIGNED: = 325, 127 SCD). Diffusion MRI was processed to generate regional and voxel-wise diffusion tensor values in bilateral lower cingulum white matter, while T1-weighted MRI was processed to generate regional volume and cortical thickness in the medial temporal lobe white matter, entorhinal cortex, temporal pole, and hippocampus. RESULTS/UNASSIGNED:SCD participants had thinner cortex in bilateral entorhinal cortex and right temporal pole. No between-group differences were noted for any of the microstructural metrics of the lower cingulum. However, correlations with delayed story recall were significant for all diffusion microstructure metrics in the right lower cingulum in SCD, but not in controls, with a significant interaction effect. Additionally, the SCD group showed an accelerated aging effect in bilateral lower cingulum with MD, AxD, and RD. DISCUSSION/UNASSIGNED:The diffusion profiles observed in both interaction effects are suggestive of a mixed neuroinflammatory and neurodegenerative pathology. Left entorhinal cortical thinning correlated with decreased FA and increased RD, suggestive of demyelination. However, right entorhinal cortical thinning also correlated with increased AxD, suggestive of a mixed pathology. This may reflect combined pathologies implicated in early AD. DTI was more sensitive than cortical thickness to the associations between SCD, memory, and age. The combined effects of mixed pathology may increase the sensitivity of DTI metrics to variations with age and cognition.

BACKGROUND AND PURPOSE/OBJECTIVE:The purpose of this study is to evaluate the feasibility of using 3-dimensional (3D) ultra-short echo time (UTE) radial imaging method for measurement of the permeability of the blood-brain barrier (BBB) to gadolinium-based contrast agent. In this study, we propose to use the golden-angle radial sparse parallel (GRASP) method with 3D center-out trajectories for UTE, hence named as 3D UTE-GRASP. We first examined the feasibility of using 3D UTE-GRASP dynamic contrast-enhanced (DCE)-magnetic resonance imaging (MRI) for differentiating subtle BBB disruptions induced by focused ultrasound (FUS). Then, we examined the BBB permeability changes in Alzheimer's disease (AD) pathology using Alzheimer's disease transgenic mice (5xFAD) at different ages. METHODS:For FUS experiments, we used four Sprague Dawley rats at similar ages where we compared BBB permeability of each rat receiving the FUS sonication with different acoustic power (0.4-1.0 MPa). For AD transgenic mice experiments, we included three 5xFAD mice (6, 12, and 16 months old) and three wild-type mice (4, 8, and 12 months old). RESULTS:The result from FUS experiments showed a progressive increase in BBB permeability with increase of acoustic power (p < .05), demonstrating the sensitivity of DCE-MRI method for detecting subtle changes in BBB disruption. Our AD transgenic mice experiments suggest an early BBB disruption in 5xFAD mice, which is further impaired with aging. CONCLUSION/CONCLUSIONS:The results in this study substantiate the feasibility of using the proposed 3D UTE-GRASP method for detecting subtle BBB permeability changes expected in neurodegenerative diseases, such as AD.

PURPOSE:In Dynamic contrast-enhanced MRI (DCE-MRI), Arterial Input Function (AIF) has been shown to be a significant contributor to uncertainty in the estimation of kinetic parameters. This study is to assess the feasibility of using a deep learning network to estimate local Capillary Input Function (CIF) to estimate blood-brain barrier (BBB) permeability, while reducing the required scan time. MATERIALS AND METHOD:-10min methods in estimating the PS values. RESULTS:-10min. We found a 75% increase of BBB permeability in the gray matter and a 35% increase in the white matter, when comparing the older group to the younger group. CONCLUSIONS:We demonstrated the feasibility of estimating the capillary-level input functions using a deep learning network. We also showed that this method can be used to estimate subtle age-related changes in BBB permeability with reduced scan time, without compromising accuracy. Moreover, the trained deep learning network can automatically select CIF, reducing the potential uncertainty resulting from manual user-intervention.

OBJECTIVES:Subjective cognitive decline (SCD) is a preclinical stage of AD. White matter hyperintensities (WMH), an MRI marker of cerebral small vessel disease, associate with AD biomarkers and progression. The impact of WMH on SCD phenotype is unclear. METHODS/DESIGN:A retrospective, cross-sectional analysis was conducted on a diverse cohort with SCD evaluated at the NYU Alzheimer's Disease Research Center between January 2017 and November 2021 (n = 234). The cohort was dichotomized into none-to-mild (n = 202) and moderate-to-severe (n = 32) WMH. Differences in SCD and neurocognitive assessments were evaluated via Wilcoxon or Fisher exact tests, with p-values adjusted for demographics using multivariable logistic regression. RESULTS:Moderate-to-severe WMH participants reported more difficulty with decision making on the Cognitive Change Index (1.5 SD 0.7 vs. 1.2 SD 0.5, p = 0.0187) and worse short-term memory (2.2 SD 0.4 vs. 1.9 SD 0.3, p = 0.0049) and higher SCD burden (9.5 SD 1.6 vs. 8.7 SD 1.7, p = 0.0411) on the Brief Cognitive Rating Scale. Moderate-to-severe WMH participants scored lower on the Mini-Mental State Examination (28.0 SD 1.6 vs. 28.5 SD 1.9, p = 0.0491), and on delayed paragraph (7.2 SD 2.0 vs. 8.8 SD 2.9, p = 0.0222) and designs recall (4.5 SD 2.3 vs. 6.1 SD 2.5, p = 0.0373) of the Guild Memory Test. CONCLUSIONS:In SCD, WMH impact overall symptom severity, specifically in executive and memory domains, as well as objective performance on global and domain-specific tests in verbal memory and visual working/associative memory.

INTRODUCTION/BACKGROUND:Mild to moderate exercise may decrease Alzheimer's disease (AD) risk, but the effects of vigorous, regular physical exercise remain unclear. METHODS:Two patients with initial diagnoses of amnestic mild cognitive impairment (MCI) demonstrated positive AD biomarkers throughout 16 and 8 years of follow-up, with final diagnoses of mild AD and amnestic MCI, respectively. RESULTS:Patient 1 was diagnosed with amnestic MCI at age 64. Neuropsychological testing, magnetic resonance imaging (MRI), fluorodeoxyglucose-positron emission tomography (FDG-PET), amyloid imaging PET, and cerebrospinal fluid (CSF) biomarkers during follow-ups remained consistent with AD. By age 80, progression was minimal with Montreal Cognitive Assessment (MoCA) 26 of 30. Patient 2 was diagnosed with amnestic MCI at age 72. Neuropsychological testing, MRI, FDG-PET, and amyloid imaging PET during follow-ups remained consistent with AD. At age 80, MoCA was 27 of 30 with no clinical progression. Both patients regularly performed vigorous, regular exercise that increased after retirement/work reduction. DISCUSSION/CONCLUSIONS:Vigorous, regular exercise may slow disease progression in biomarker-positive amnestic MCI and mild AD.