Related Experiment Video
Updated: Jun 4, 2026
Meta-analysis of Voxel-Based Neuroimaging Studies using Seed-based d Mapping with Permutation of Subject Images (SDM-PSI)
Published on: November 27, 2019
Bias in tensor based morphometry Stat-ROI measures may result in unrealistic power estimates
Wesley K Thompson1, Dominic Holland2,
1Department of Psychiatry, The University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92037, USA; Stein Institute for Research on Aging, The University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92037, USA.
This study reveals a methodological bias in Alzheimer's Disease Neuroimaging Initiative (ADNI) data analysis. Early rapid atrophy in Alzheimer's disease and mild cognitive impairment subjects suggests potential overestimation of disease progression rates.
Area of Science:
- Neuroimaging
- Neurology
- Biostatistics
Background:
- Recent studies utilize tensor-based morphometry and statistically-defined regions of interest (Stat-ROIs) to analyze longitudinal brain atrophy in Alzheimer's Disease Neuroimaging Initiative (ADNI) structural MRIs.
- This commentary critically examines Hua et al. (2010) and related reports, addressing the broader implications for Alzheimer's disease (AD) and mild cognitive impairment (MCI) research.
Purpose of the Study:
- To identify and address potential methodological biases in the quantification of longitudinal brain atrophy using tensor-based morphometry in ADNI data.
- To evaluate the temporal pattern of atrophy in AD, MCI, and healthy control subjects within the ADNI cohort.
Main Methods:
- Analysis of publicly available ADNI structural MRI data.
- Application of tensor-based morphometry to quantify regional brain volume changes over time.
- Comparison of atrophy patterns between Alzheimer's disease, mild cognitive impairment, and healthy control groups.
Main Results:
- A significant temporal pattern of atrophy was observed, with the majority of volume loss occurring within the first six months of a two-year period.
- This pattern was present in both AD and MCI subjects, as well as in identically processed healthy controls.
- The findings strongly suggest a methodological bias corrupting the atrophy measures.
Conclusions:
- The identified methodological bias significantly impacts the validity of conclusions drawn from studies using these measures.
- Estimated rates of change in brain atrophy may be artificially elevated, particularly in the early stages of the disease.
- Sample size estimations, such as those reported by Hua et al. (2010), could be substantially underestimated, potentially by a factor of five to sixteen.
