Towards a multimodal neuroimaging-based risk score for mild cognitive impairment by combining clinical studies with a large (N>37000) population-based study

Elaheh Zendehrouh, Mohammad S.e. Sendi, Anees Abrol, Ishaan Batta, Reihaneh Hassanzadeh, Vince D. CalhounmedRxiv 2024 · DOI

TL;DR

Researchers have developed a "Brain-wide Risk Score" (BRS) that uses non-invasive structural and functional MRI data to identify individuals at high risk for Mild Cognitive Impairment (MCI). By validating this score across three major datasets—including the massive UK Biobank (N > 37,000)—the study demonstrates that combining gray matter patterns with functional connectivity can effectively flag early neurodegenerative changes before a clinical diagnosis is made.

What Problem Does This Paper Address?

Alzheimer’s Disease (AD) is a global health crisis, but by the time clinical symptoms are obvious, significant brain damage has often already occurred. The stage just before AD, known as Mild Cognitive Impairment (MCI), is the critical window for intervention.

Currently, risk assessment relies heavily on genetics (which only tells part of the story) or invasive procedures like spinal taps (CSF analysis) and radioactive PET scans. There is a pressing need for a non-invasive, imaging-based biomarker that can capture the complex interplay of genetics, environment, and lifestyle. This paper tackles the challenge of creating a reliable "risk score" using standard MRI scans that can be applied to the general population.

Methods

The authors introduced a pipeline to calculate the Brain-wide Risk Score (BRS) based on two neuroimaging modalities:

  1. Structural MRI (sMRI): Used to create Gray Matter (GM) segmentation maps, looking for patterns of tissue loss.
  2. Resting-state Functional MRI (fMRI): Used to calculate Functional Network Connectivity (FNC) matrices, which describe how different brain regions communicate.

The BRS Logic: The researchers first established "reference patterns" for Healthy Controls (CN) and MCI patients using the OASIS-3 dataset. For any new participant, they calculated two correlation values: how much the participant's brain looks like a healthy brain ($R_{CN}$) and how much it looks like an MCI brain ($R_{MCI}$). The final score is simply the difference: $BRS = R_{CN} - R_{MCI}$. A lower (or negative) score indicates a brain pattern more similar to MCI.

They used a "NeuroMark" pipeline for fMRI, which employs spatially constrained Independent Component Analysis (ICA) to ensure that functional networks are identified consistently across different subjects and datasets.

Datasets

The study leveraged three major neuroimaging repositories:

  • OASIS-3: Used as the primary reference dataset (1,028 healthy controls, 361 MCI).
  • ADNI (Alzheimer’s Disease Neuroimaging Initiative): Used as an independent validation set to ensure the BRS wasn't just "overfitting" to one group.
  • UK Biobank (UKB): The target population-based study, featuring over 37,000 participants. This allowed the researchers to see how the risk score distributes across a large, mostly healthy aging population.

Key Insights and Ideas

  • Structural vs. Functional: While both modalities worked, structural MRI (Gray Matter) was a stronger differentiator between healthy aging and MCI than functional connectivity. This aligns with the idea that physical atrophy is a late-stage but very clear marker of decline.
  • The Power of Multimodality: The most robust results came from combining both scores. Some individuals showed high risk in their brain structure but not function (or vice versa), suggesting these two markers provide complementary information about brain health.
  • Sensory Networks Matter: Interestingly, the researchers found that healthy controls showed much higher connectivity in sensory networks (visual, auditory, motor) compared to the MCI group. In MCI, these networks often showed highly negative connectivity, suggesting that sensory-motor degradation might be an overlooked early sign of cognitive decline.
  • Generalizability: The BRS worked even when the "reference" was swapped. Whether using OASIS-3 or ADNI as the baseline, the score consistently identified the same high-risk individuals in the UK Biobank.

Key Results

  • Statistical Significance: In the OASIS-3 dataset, the BRS showed massive separation between groups, with p-values as low as $3.4 \times 10^{-35}$ for structural MRI.
  • Validation: When applied to the independent ADNI dataset, the BRS successfully differentiated MCI from controls ($p = 1.9 \times 10^{-14}$ for the combined multimodal score).
  • Real-world Efficacy: Within the UK Biobank, the researchers identified 46 participants who were officially diagnosed with Alzheimer’s within five years after their initial scan. These individuals' BRS scores consistently placed them in the "high-risk" subgroups identified by the model, proving the score's predictive potential.
  • Subgrouping: The team stratified the 37,000+ UKB participants into 10 "risk deciles." They found that the highest-risk decile (Subgroup 1) almost perfectly mirrored the brain patterns of clinically diagnosed MCI patients.

Limitations and Future Work

The authors acknowledge that using the "whole brain" for the score might introduce noise, as some regions are less relevant to Alzheimer’s than others. Future iterations might focus specifically on high-impact areas like the hippocampus or sensory-motor networks to sharpen the score's accuracy.

Additionally, the study did not directly compare the BRS to Polygenic Risk Scores (PRS). A major next step will be to see if combining imaging-based risk with genetic risk provides an even more powerful tool for early detection. Finally, the researchers suggest exploring different "distance metrics" beyond simple correlation to see if more complex mathematical models can better capture the nuances of brain aging.

Summary updated

Abstract

Abstract Alzheimer’s disease (AD) is the most common form of age-related dementia, leading to a decline in memory, reasoning, and social skills. While numerous studies have investigated the genetic risk factors associated with AD, less attention has been given to identifying a brain imaging-based measure of AD risk. This study introduces a novel approach to assess mild cognitive impairment MCI, as a stage before AD, risk using neuroimaging data, referred to as a brain-wide risk score (BRS), which incorporates multimodal brain imaging. To begin, we first categorized participants from the Open Access Series of Imaging Studies (OASIS)-3 cohort into two groups: controls (CN) and individuals with MCI. Next, we computed structure and functional imaging features from all the OASIS data as well as all the UK Biobank data. For resting functional magnetic resonance imaging (fMRI) data, we computed functional network connectivity (FNC) matrices using fully automated spatially constrained independent component analysis. For structural MRI data we computed gray matter (GM) segmentation maps. We then evaluated the similarity between each participant’s neuroimaging features from the UK Biobank and the difference in the average of those features between CN individuals and those with MCI, which we refer to as the brain-wide risk score (BRS). Both GM and FNC features were utilized in determining the BRS. We first evaluated the differences in the distribution of the BRS for CN vs MCI within the OASIS-3 (using OASIS-3 as the reference group). Next, we evaluated the BRS in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort (using OASIS-3 as the reference group), showing that the BRS can differentiate MCI from CN in an independent data set. Subsequently, using the sMRI BRS, we identified 10 distinct subgroups and similarly, we identified another set of 10 subgroups using the FNC BRS. For sMRI and FNC we observed results that mutually validate each other, with certain aspects being complementary. For the unimodal analysis, sMRI provides greater differentiation between MCI and CN individuals than the fMRI data, consistent with prior work. Additionally, by utilizing a multimodal BRS approach, which combines both GM and FNC assessments, we identified two groups of subjects using the multimodal BRS scores. One group exhibits high MCI risk with both negative GM and FNC BRS, while the other shows low MCI risk with both positive GM and FNC BRS. Moreover, in the UKBB we have 46 participants diagnosed with AD showed FNC and GM patterns similar to those in high-risk groups, defined in both unimodal and multimodal BRS. Finally, to ensure the reproducibility of our findings, we conducted a validation analysis using the ADNI as an additional reference dataset and repeated the above analysis. The results were consistently replicated across different reference groups, highlighting the potential of FNC and sMRI-based BRS in early Alzheimer’s detection.
BibTeX
@misc{zendehrouh2024towards,
  author = {Elaheh Zendehrouh and Mohammad S.E. Sendi and Anees Abrol and Ishaan Batta and Reihaneh Hassanzadeh and Vince D. Calhoun},
  title = {{Towards a multimodal neuroimaging-based risk score for mild cognitive impairment by combining clinical studies with a large (N\>37000) population-based study}},
  year = {2024},
  doi = {10.1101/2024.03.12.24303873},
  howpublished = {medRxiv},
}
Figure