There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | Xiaojing Li |
| Institution: | Carl von Ossietzky Universität Oldenburg |
| Department: | psychology |
| Country: | |
| Proposed Analysis: | While prediction of Alzheimer’s Dementia (AD) symptoms onset based on biomarkers (including genetics) and cognitive and non-cognitive reserve factors has been the focus of many studies using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) databases, only a few studies aimed to predict and understand the disease progress trajectory before and after symptoms onset (e.g., Melis et al., 2019, Curr Opin Psychiatr). However, the prediction of such trajectories are key for future intervention planning and for providing supportive information for clinicians and caregivers. Hereby we request data of the three ADNI cohorts in order to achieve the following aims: 1. Model the longitudinal trajectory of symptoms progress for the time frame of pre and post AD diagnosis; 2. Identify latent classes of trajectory types; 3. Use biomarkers and reserve factors to predict trajectory types. In order to achieve these aims, we would model the data of healthy control, MCI and AD individuals belonging to the three ADNI cohorts ADNI1, ADNI2 and ADNI3. Latent growth curve mixture models will be used to identify the trajectory classes based on the following measurements at all available measurement time points: • As diagnosis instruments of AD we aim to use ADAS-Cog, MMSE, MoCA and CDR scores (including global and subscale values to differentiate cognitive and daily life functioning); • As cognitive ability measures we would include the ANART, the Clock drawing test, Logical memory, Category fluency, Boston naming test, Rey Auditory Verbal Learning Test and Trail Making Test. Determined trajectories based on both these classes of measurement instruments will be jointly modelled for class identification. Thus, multiple tests will be used to identify time point specific latent variables representing AD symptoms of daily life functioning decline and cognitive performance factors, so that trajectories can be established at the latent, measurement-error-free level. Furthermore, there are a few first attempts to identify factors for prediction of AD disease progress (e.g., Melis et al., 2019, Curr Opin Psychiatr). Theoretically conceivable factors belong to multiple domains, such as disease syndrome level and comorbidities, neuropsychological functioning and neuropsychiatric symptoms, blood and imaging based biomarkers, genetic predispositions, frailty levels, physical activity and socio-demographics, some of which are considered so called reserve factors in the literature (e.g., Ewers, 2020, Curr Opin Psychiatr). Hitherto exiting studies on trajectory prediction mainly focused on isolated factor bundles from the above. However, future studies should go beyond these first attempts by jointly modeling multiple factors present at baseline or such that are time varying along the trajectories. Thus, our aim is to jointly model time invariant and time varying predictors of AD disease progression, by including the following: • Genetic risk factors, focusing especially on the APOE polymorphism; • Time varying structural brain characteristics including cortical thinning at baseline and its development over time, hippocampus atrophy, white matter / grey matter density change and white matter lesions; • Time varying resting state connectivity and dispersion entropy in the fronto-parietal, the salience and the default mode network, • Neuropsychological functioning and neuropsychiatric symptoms, including everyday cognition (ECOG), emotion (NPI-Q), depression at baseline and depression development over time (GDS), as well as instrumental activities of daily living (IDAL) • Socio-demographic factors and physical fitness, more specifically education, gender, medical history, age at first symptom onset and vital signs over time, including pulse rate, temperature, respiration rate, and blood pressure, that indicate the state of a patient's essential body functions. For prediction of trajectory type class membership, we will explore the classification accuracy of linear and quadratic discriminant analysis methods, classification trees and random forests, as well as support vector machines. For assessing predictor utility, we will apply currently proposed indices for understandable machine learning. We believe that the joint modelling of these predictors at the time invariant and time variant level, as well as the joint trajectory modeling of cognitive and non-cognitive symptoms for latent class identification will shed new light on understanding AD disease progression to the realm of better future intervention planning. |
| Additional Investigators | |
| Investigator's Name: | Andrea Hildebrandt |
| Proposed Analysis: | While prediction of Alzheimer’s Dementia (AD) symptoms onset based on biomarkers (including genetics) and cognitive and non-cognitive reserve factors has been the focus of many studies using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) databases, only a few studies aimed to predict and understand the disease progress trajectory before and after symptoms onset (e.g., Melis et al., 2019, Curr Opin Psychiatr). However, the prediction of such trajectories are key for future intervention planning and for providing supportive information for clinicians and caregivers. Hereby we request data of the three ADNI cohorts in order to achieve the following aims: 1. Model the longitudinal trajectory of symptoms progress for the time frame of pre and post AD diagnosis; 2. Identify latent classes of trajectory types; 3. Use biomarkers and reserve factors to predict trajectory types. In order to achieve these aims, we would model the data of healthy control, MCI and AD individuals belonging to the three ADNI cohorts ADNI1, ADNI2 and ADNI3. Latent growth curve mixture models will be used to identify the trajectory classes based on the following measurements at all available measurement time points: • As diagnosis instruments of AD we aim to use ADAS-Cog, MMSE, MoCA and CDR scores (including global and subscale values to differentiate cognitive and daily life functioning); • As cognitive ability measures we would include the ANART, the Clock drawing test, Logical memory, Category fluency, Boston naming test, Rey Auditory Verbal Learning Test and Trail Making Test. Determined trajectories based on both these classes of measurement instruments will be jointly modelled for class identification. Thus, multiple tests will be used to identify time point specific latent variables representing AD symptoms of daily life functioning decline and cognitive performance factors, so that trajectories can be established at the latent, measurement-error-free level. Furthermore, there are a few first attempts to identify factors for prediction of AD disease progress (e.g., Melis et al., 2019, Curr Opin Psychiatr). Theoretically conceivable factors belong to multiple domains, such as disease syndrome level and comorbidities, neuropsychological functioning and neuropsychiatric symptoms, blood and imaging based biomarkers, genetic predispositions, frailty levels, physical activity and socio-demographics, some of which are considered so called reserve factors in the literature (e.g., Ewers, 2020, Curr Opin Psychiatr). Hitherto exiting studies on trajectory prediction mainly focused on isolated factor bundles from the above. However, future studies should go beyond these first attempts by jointly modeling multiple factors present at baseline or such that are time varying along the trajectories. Thus, our aim is to jointly model time invariant and time varying predictors of AD disease progression, by including the following: • Genetic risk factors, focusing especially on the APOE polymorphism; • Time varying structural brain characteristics including cortical thinning at baseline and its development over time, hippocampus atrophy, white matter / grey matter density change and white matter lesions; • Time varying resting state connectivity and dispersion entropy in the fronto-parietal, the salience and the default mode network, • Neuropsychological functioning and neuropsychiatric symptoms, including everyday cognition (ECOG), emotion (NPI-Q), depression at baseline and depression development over time (GDS), as well as instrumental activities of daily living (IDAL) • Socio-demographic factors and physical fitness, more specifically education, gender, medical history, age at first symptom onset and vital signs over time, including pulse rate, temperature, respiration rate, and blood pressure, that indicate the state of a patient's essential body functions. For prediction of trajectory type class membership, we will explore the classification accuracy of linear and quadratic discriminant analysis methods, classification trees and random forests, as well as support vector machines. For assessing predictor utility, we will apply currently proposed indices for understandable machine learning. We believe that the joint modelling of these predictors at the time invariant and time variant level, as well as the joint trajectory modeling of cognitive and non-cognitive symptoms for latent class identification will shed new light on understanding AD disease progression to the realm of better future intervention planning. |
