ADNI data is made available to researchers around the world. As such, there are many active research projects accessing and applying the shared ADNI data. To further encourage Alzheimer’s disease research collaboration, and to help prevent duplicate efforts, the list below shows the specific research focus of the active ADNI investigations. This information is requested annually as a requirement for data access.
|Principal Investigator's Name:||Dimitrios Kapogiannis|
|Institution:||National Institute on Aging|
|Department:||Clinical Research Branch|
|Proposed Analysis:||OBJECTIVE To use Source Based Morphometry (SBM), a hypothesis-free methodology for structural MRI analysis conceptually similar to group Independent Component Analysis (Group ICA) for fMRI, on longitudinally collected 3T MP-RAGE images of control, MCI, and AD participants in ADNI. The outcome measure of interest would be spatial maps representing orthogonal patterns of volumetric atrophy. CSF biomarkers, specifically Aβ1-42, Tau, and P-tau181P, and plasma Aβ1-40, Aβ1-42 would be used to predict atrophy represented by these spatial component maps between groups. This approach could define differential regional patterns of volumetric change associated with Aβ and tau. BACKGROUND The assessment of volumetric atrophy in MR images typically relies on deformation- or classifier-based data parameterization using a general linear model (GLM). In ADNI, Thompson’s group and others have elegantly shown that tensor based morphometry (TBM) can produce 3-Dimensional Jacobian maps that represent single-subject tissue expansion and contraction throughout the brain. This whole brain Jacobian map can then be used, in voxel-wise analyses, to determine whether or not a group classification or predictor significantly explains regional atrophy in a priori areas of interest (Hua et al., 2009; Wang et al., 2011). By contrast, machine learning feature selection and linear support vector machines (SVMs) can classify voxels into meaningful spatial patterns (Sahiner et al., 2000; Bishop, 2006). SVMs can be trained on imaging datasets to, for example, distinguish AD versus control brains (Klöppel et al., 2008). Although this machine learning approach could, in theory, determine patterns of atrophy that are orthogonal among groups in a data-driven fashion, the SVM must always be trained on datasets based on a priori classification of how that data should be binned. It is therefore of interest to see if a purely data-driven approach, such as SBM, can extract different spatiotemporal patterns of atrophy across time without a priori parameterization, which may be more sensitive than TBM or SVMs. As explained by Calhoun and colleagues (2008), Group ICA (and SBM) generate(s) orthogonalized spatial or temporal patterns from time-course data by: 1) principle component analysis (PCA) data reduction on the subject level; 2) PCA on the group level; 3) ICA to achieve component factor independence based on higher order statistics; and 4) back-reconstruction of components onto subject image data for use in a GLM. This technique has been used thus far on fMRI, EEG, or other 4-Dimensional data collected in a single recording session. In a pilot analysis using structural MRI data from the Baltimore Longitudinal Study of Aging, we found that SBM algorithms can be used to analyze longitudinally collected T1 images, producing meaningful spatial components. Specifically, these spatial maps bear a strong resemblance to the typical tissue expansion and contraction result maps generated when regressing age against TBM-derived Jacobian images. We would be able to better assess this methodology through the use of a larger longitudinal dataset such as ADNI. In addition, we believe that the greatest potential of this methodology lies in associating Independent Components with CSF and plasma biomarkers collected in ADNI. DATA REQUEST For MR data, we would like to download 3T MP-RAGE images longitudinally acquired in control, MCI, and AD patients cross-listed as having Lumbar Puncture (LP) data at Baseline. Specifically, according to the LONI website, we would like to examine images collected at Baseline, Month 6, Month 12, and Month 24. For CSF data, we would like to download the biomarkers Aβ1-42, Tau, and P-tau181P from Baseline; for plasma, we would like to download Aβ1-40 and Aβ1-42 data. For clinical data, we wish to access: 1) demographic information from Screening; and 2) neuropsychological data from Baseline, Month 6, Month 12, and Month 24 for potential mediational analyses (see Willette et al., in press). APPROACH SBM, a toolbox in SPM8 (http://mialab.mrn.org/software/gift/), would be used to perform Group ICA in structural MRI data. Longitudinal scans for control, MCI, and AD participants would be inputted. The Infomax ICA algorithm would be used to derive independent components explaining spatiotemporal variance across the 4 timepoints. Ideally, these components would capture in a purely data-driven fashion differential patterns of atrophy related to aging, the amyloid cascade (Aβ), and neurodegeneration (tau). GICA3-based back-reconstruction would then be used to derive spatial component maps of atrophy specific to each participant. A multiple regression model in SPM8 would be created. The dependent variable would be subject-specific spatial maps for a given component. The predictor would be CSF and plasma biomarker data such as P-tau181P or the P-tau181P/Aβ1–42 ratio (see below). Initial covariates would include age and sex. Voxel and cluster p values would be .001 (uncorrected) and .05 (corrected) respectively, with cluster correction done using Monte Carlo simulations in AlphaSim. Permutational testing to derive an appropriate statistical threshold was also used by Leow and colleague (2009) in a paper we will emulate as discussed below. EXPECTED OUTCOMES Based on biomarker results in predicting TBM-related atrophy in ADNI (Leow et al., 2009), we hypothesize that: 1) higher P-tau181P or a higher P-tau181P/Aβ1–42 ratio would not be associated with regional atrophy in the components representing expansion and contraction typical of normative aging; 2) higher P-tau181P or a higher P-tau181P/Aβ1–42 ratio would significantly predict greater anterior medial and inferior temporal atrophy in the MCI-like cluster; 3) higher P-tau181P or a higher P-tau181P/Aβ1–42 ratio would significantly predict atrophy in medial temporal, prefrontal, and parietal gyri in the AD-like cluster. SUMMARY A limitation of TBM and SVMs is that GLM analyses test for regional differences among voxels based on a priori selection. The advantage of Group ICA is that patterns of volumetric atrophy may be extracted from the longitudinal data in a data-driven manner with fewer assumptions. The spatial components generated by Group ICA may resemble atrophy patterns reminiscent of the three groups examined. SPM-based GLMs using biomarker data would allow us to confirm if these patterns of atrophy are modulated by P-tau181P in a manner comparable to Jacobian maps used by Leow and colleagues (2009). We would thus be able to validate if our component-based outcome measure may be useful in other paradigms. REFERENCES Bishop C. Pattern Recognition and Machine Learning. Springer; New York: 2006. Calhoun VD, Liu J, Adali T. A review of group ICA for fMRI data and ICA for joint inference of imaging, genetic, and ERP data. Neuroimage 2009; 45(1 Suppl): S163-72. Hua X, Leow AD, Parikshak N, Lee S, Chiang MC, Toga AW, Jack CR Jr, Weiner MW, Thompson PM; Alzheimer's Disease Neuroimaging Initiative. Tensor-based morphometry as a neuroimaging biomarker for Alzheimer's disease: an MRI study of 676 AD, MCI, and normal subjects. Neuroimage 2008; 43(3): 458-69 Klöppel S, Stonnington C, Chu C, Draganski B, Scahill R, Rohrer J, Fox N, Jack C, Ashburner J, Frackowiak R. Automatic classification of MR scans in Alzheimer's disease. Brain 2008; 131(3): 681–689. Leow AD, Yanovsky I, Parikshak N, Hua X, Lee S, Toga AW, Jack CR Jr, Bernstein MA, Britson PJ, Gunter JL, Ward CP, Borowski B, Shaw LM, Trojanowski JQ, Fleisher AS, Harvey D, Kornak J, Schuff N, Alexander GE, Weiner MW, Thompson PM; Alzheimer's Disease Neuroimaging Initiative. Alzheimer's disease neuroimaging initiative: a one-year follow up study using tensor-based morphometry correlating degenerative rates, biomarkers and cognition. Neuroimage 2009; 45(3): 645-55. Sahiner B, Chan H, Petrick N, Wagner R, Hadjiiski L. Feature selection and classifier performance in computer-aided diagnosis: the effect of finite sample size. Med. Phys. 2000; 27(7): 1509–1522. Wang Y, Song Y, Rajagopalan P, An T, Liu K, Chou YY, Gutman B, Toga AW, Thompson PM; Alzheimer's Disease Neuroimaging Initiative. Surface-based TBM boosts power to detect disease effects on the brain: an N=804 ADNI study. Neuroimage 2011; 56(4): 1993-2010. Willette AA, Bendlin BB, Colman RJ, Kastman EK, Field AS, Alexander AL, Sridharan A, Allison DB, Anderson R, Voytko ML, Kemnitz JW, Weindruch RH, Johnson SC. Calorie Restriction Reduces the Influence of Glucoregulatory Dysfunction on Regional Brain Volume in Aged Rhesus Monkeys. Diabetes 2012. In press.|
|Investigator's Name:||Auriel Willette|
|Proposed Analysis:||Drs. Kapogiannis and Willette are jointly proposing the same analysis to ADNI for consideration. Please see the proposal listed under Dr. Kapogiannis.|