The Harvard Aging Brain Study (HABS) is a longitudinal observational study aimed at improving our understanding of brain aging and the earliest (preclinical) stages of Alzheimer’s disease. This study acquires imaging data to detect early brain changes related to Alzheimer’s disease, including amyloid plaques and tau tangles, as well as functional and structural imaging data, and detailed assessments of memory and other cognitive processes. Our ultimate goal is to provide knowledge that will accelerate progress towards the successful prevention of cognitive decline due to Alzheimer’s disease.
The current HABS public dataset (v2.0) includes 290 participants with longitudinal observations up to 5 years from baseline. Participants range in age from 62 to 90 years of age at baseline, and all participants were deem non-clinically impaired at baseline. Clinical and cognitive measures are typically collected annually, while imaging measures are typically collected at baseline, at 3 years, and at 5 years.
Participants were enrolled in the HABS cohort if they were deemed cognitively normal based on a CDR evaluation and performance on the Logical Memory delayed recall score, MMSE, and GDS.1,3,4 All clinical, neuropsychological, and imaging assessments took place at the MGH PET Core Laboratory and the Athinoula A. Martinos Center for biomedical imaging in Boston, MA.
The HABS public dataset v2.0, contains the following data:
|Age at Enrollment|
Median [min, max]
72.1 [62.5 89.3]
75.0 [65.0 90.0]
73 [62.5, 90]
|APOE e4 status|
|Diagnosis at Last Visit|
Demographics include Time from baseline, self-reported Age, biological sex, years of education, race, Hollingshead score, and ethnicity, as well as verbal IQ (AMNART) and APOE4 status.
Clinical measures include a consensus diagnosis (cognitively normal, mild cognitive impairment, or dementia), the Clinical Dementia Rating (CDR), the MMSE, the geriatric depression scale (GDS), and the Hachinski ischemia score.
Neuropsychological assessments include the Boston Naming Task (BNT), the Categories task (CAT), the Digits task (Digits), Phonemic Fluency (FAS), the free and cued selective reminding task (FCsrt), the selective reminding task (SRT), the letter number task, the logical memory task, the trail making task (TMT), the visual form discrimination task (VFDT), the Digit Symbol task, and the Structured Telephone Interview for Dementia Assessment (DemoMemQs).
Preclinical Alzheimer’s Cognitive Composite (PACC):
We are also sharing the PACC96 and PACC5 composites, as well as the constituent normalized measurements (MMSE, Logical Memory, FCsrt96, Digits, and Categories).
A detailed description of these variables can be viewed here.
- Amyloid PET: 11C Pittsburgh Compound B (data set includes PETsurfer based ROI measurements).
- Tau PET: 18F Flortaucipir (data set includes PETsurfer based ROI measurements).
- Brain metabolism: 18F fludeoxyglucose (FDG) (data set includes PETsurfer based ROI measurements).
MRI: (sequence protocols may be viewd by clinking on the links below)
- ADNI2 T1-MPRAGE (data set includes FreeSurfer based Desikan-Killiany Atlas ROI measurements).
- ADNI2GO T1-MPRAGE (data set includes FreeSurfer based Desikan-Killiany Atlas ROI measurements).
- 30-Direction DTI
- Resting State fMRI
A detailed description of the neuroimaging data can be viewed here.
PET Processing and ROI measures.
All PET data were acquired using a Siemens/CTI ECAT HR Scanner (3D mode; 63 image planes; 15.2 cm axial field of view; 5.6 mm transaxial resolution; 2.4 mm slice interval). Before injection, 10-minute transmission scans for attenuation correction were collected. PET data were reconstructed, attenuation corrected, and then manually evaluated to verify adequate count statistics and to correct for motion.
PiB-PET: After injection of 8.5-15 mCi 11C PiB, 60-minutes of dynamic data were acquired in 3D acquisition mode and reconstructed in 39 frames (8 x 15s, 4 x 60s, and 27 x 120s).
FTP-PET: 18F Flortaucipir (FTP, aka AV1451, aka T807) was prepared at Massachusetts General Hospital with a mean radiochemical yield of 14 ± 3% and specific activity of 216 ± 60 GBq/mol (5837 ± 1621 mCi/mol) at the end of synthesis (60 min) and validated for human use (Shoup et al., 2013). After a 10.0 ± 1.0 mCi bolus injection, images were acquired from either 80 to 100 min in 4 x 5 min frames or 75 to 105 min in 6 x 5 min frames.
FDG-PET: 18F fludeoxyglucose (FDG) was acquired from 45 to 75 minutes after a 5-10mCi bolus injection and reconstructed into a single 30 minute frame.
Data Processing and ROI based measures: Briefly, following acquisition, a mean image was created (note for PiB this was the first 8 minutes post-injection), and was coregistered to the corresponding FS processed T1 image with a 6 DoF rigid body registration using spm_coreg from the SPM12 package. The gtmseg atlas was created with the gtmseg function in Freesurfer version 6, which was then used with the mri_gtmpvc function to generate both partial volume corrected (PVC) and non-PVC ROI measures within the 100 regions of interest defined in the gtmseg.mgz atlas.
mri_gtmpvc –auto-mask 11 .001 –mgx .25 –threads 8 –seg gtmseg.mgz –no-rescale –save-input –i <PETimage> –reg </pth/to/reg.lta> –psf 6 –o ‘ </pth/to/output/folder> –default-seg-merge;
Bilateral cerebellum grey matter was then used as the reference region for both SUVR and DVR measurements.
T1 MRI processing and ROI measures.
Magnetic resonance imaging (MRI) scanning was completed on a Siemens TIM Trio 3T System with a 12-channel head coil. Structural T1-weighted volumetric magnetization-prepared, rapid acquisition gradient echo (MPRAGE) scans were collected with one of two acquisitions:
- ADNI2 MPRAGE: repetition time = 2300ms, echo time = 2.98ms, inversion time = 900ms, flip angle = 9°, 1x1x1.2mm resolution, 0 acceleration
- ADNI2GO MPRAGE: repetition time = 2300ms, echo time = 2.95ms, inversion time = 900ms, flip angle = 9°, 1.1×1.1×1.2mm resolution with 2X GRAPPA acceleration
Note, these acquisitions are generally considered to be interchangeable.
Only ROI values for the ADNI MPRAGE sequences are provided. Region of interest (ROI) labeling was implemented using FreeSurfer v6.0 (http://surfer.nmr.mgh.harvard.edu/). Cortical region of interest measurements were made using the Desikan-Killiany atlas (https://surfer.nmr.mgh.harvard.edu/fswiki/CorticalParcellation). Subcortical region of interest measurements were made using the Freesurfer aseg atlas (https://surfer.nmr.mgh.harvard.edu/ftp/articles/fischl02-labeling.pdf)
Quality assurance (QA) of the ADNI style structural images involved manual assessment of the FS recons after running the full auto-recon process. This assessment included examination of the white and pial surface segmentation using the brainmask.mgz file. In cases where dura or skull influenced the segmentation result, voxels were either manually edited or corrected by adjusting the watershed threshold. In cases where the grey matter ribbon clearly included white matter or clearly excluded grey matter, control points were added to the recon and/or white matter edits were made to the wm.mgz file. autorecon2 and autorecon3 processing steps were re-run on the edited files, and the process was repeated until the segmentation results were deemed either sufficient or irreparable.