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: | Jessica Pommy |
| Institution: | Medical College of Wisconsin |
| Department: | Neuropsychology |
| Country: | |
| Proposed Analysis: | Neurodegenerative disorders are characterized by heterogeneity in terms of clinical presentation, etiology, and course. Broadly, the goal of the proposed study is to characterize this heterogeneity in a meaningful way. A variety of different approaches have been used to identify mild cognitive impairment (MCI) phenotypes within heterogenous samples, including hierarchical cluster analysis, latent factor analysis, etc. One limitation of these approaches is the user typically must identify the number of clusters a priori. We propose using a data-driven approach, community detection (CD), to identify MCI phenotypes using ADNI data. CD is an analytic approach based on graph theory that examines modularity within a sample (Newman, 2006). This approach has traditionally been applied to analyses of neural networks and social networks. However, previous work by Fair and colleagues has demonstrated the utility of this approach when examining neuropsychological heterogeneity, we propose CD may be a novel method for examining heterogeneity associated with aging and neurodegenerative processes at the MCI stage (Fair et al., 2012). We then propose to examine the utility of the cognitive subtypes identified with CD using a machine learning approach as was demonstrated by Fair and colleagues. More specifically, we would assess whether community membership increases the accuracy of a support vector machine (SVM) based multivariate pattern analysis (MVPA) classifier in predicting group membership (MCI versus Healthy Control) using neuropsychological test scores. Lastly, we propose follow up exploratory analyses to examine potential correlates associated with membership to a specific community by examining clinical characteristics and neuroimaging and other neurodegenerative biomarkers (e.g., APOE status, Ab42/40, PET, etc.) between MCI phenotypes (e.g., examining if vascular risk factors are higher in one subtype or hippocampal and other atrophy patterns are associated with given subtype(s)). Ultimately, if we are able to detect and validate this approach in the ADNI dataset, we would like to attempt to replicate these findings in a clinical sample (N = 150) diagnosed with amnestic or nonamnestic MCI that we have collected at our site with approval from our IRB. We have collected both comprehensive neuropsychological data and neuroimaging data on our sample. We believe the proposed analyses will be meaningful because 1) we can examine heterogeneity in both healthy controls and MCI samples, 2) community subtypes may represent underlying endophenotypes relevant for MCI, 3) we can determine the added value of MRI biomarkers to identification of CD-based MCI phenotypes and compare to cluster analytic approaches. |
| Additional Investigators | |
| Investigator's Name: | Laura Umfleet |
| Proposed Analysis: | Neurodegenerative disorders are characterized by heterogeneity in terms of clinical presentation, etiology, and course. Broadly, the goal of the proposed study is to characterize this heterogeneity in a meaningful way. A variety of different approaches have been used to identify mild cognitive impairment (MCI) phenotypes within heterogenous samples, including hierarchical cluster analysis, latent factor analysis, etc. One limitation of these approaches is the user typically must identify the number of clusters a priori. We propose using a data-driven approach, community detection (CD), to identify MCI phenotypes using ADNI data. CD is an analytic approach based on graph theory that examines modularity within a sample (Newman, 2006). This approach has traditionally been applied to analyses of neural networks and social networks. However, previous work by Fair and colleagues has demonstrated the utility of this approach when examining neuropsychological heterogeneity, we propose CD may be a novel method for examining heterogeneity associated with aging and neurodegenerative processes at the MCI stage (Fair et al., 2012). We then propose to examine the utility of the cognitive subtypes identified with CD using a machine learning approach as was demonstrated by Fair and colleagues. More specifically, we would assess whether community membership increases the accuracy of a support vector machine (SVM) based multivariate pattern analysis (MVPA) classifier in predicting group membership (MCI versus Healthy Control) using neuropsychological test scores. Lastly, we propose follow up exploratory analyses to examine potential correlates associated with membership to a specific community by examining clinical characteristics and neuroimaging and other neurodegenerative biomarkers (e.g., APOE status, Ab42/40, PET, etc.) between MCI phenotypes (e.g., examining if vascular risk factors are higher in one subtype or hippocampal and other atrophy patterns are associated with given subtype(s)). Ultimately, if we are able to detect and validate this approach in the ADNI dataset, we would like to attempt to replicate these findings in a clinical sample (N = 150) diagnosed with amnestic or nonamnestic MCI that we have collected at our site with approval from our IRB. We have collected both comprehensive neuropsychological data and neuroimaging data on our sample. We believe the proposed analyses will be meaningful because 1) we can examine heterogeneity in both healthy controls and MCI samples, 2) community subtypes may represent underlying endophenotypes relevant for MCI, 3) we can determine the added value of MRI biomarkers to identification of CD-based MCI phenotypes and compare to cluster analytic approaches. |
