A study of 1,462 healthy adults linked MRI-based brain 'age' to memory, mood, and gut bacteria
A study of 1,462 healthy adults linked MRI-based brain 'age' to memory, mood, and gut bacteria
On September 9, the journal eBioMedicine published a paper on healthy individuals aged 18 to 65. The authors trained a model to estimate age from resting-state functional MRI, which measures how signals across different brain regions rise and fall in synchrony while a person lies in the scanner. They validated this estimate in 1,462 participants across three cohorts, and in 182 of them compared it with bacterial DNA and molecules from stool samples.
Resting-state functional MRI records signal fluctuations across many brain regions and allows one to calculate which pairs of regions change together. The model was trained on data from 674 individuals, then tested on additional groups of 444 and 344. In 40 participants, repeated scans on the same day produced similar estimates.
To make people of different ages comparable, the authors removed the systematic bias linked to chronological age from the prediction. The remaining difference between the model's estimate and calendar age became the BAI index. A positive value means that, based on the pattern of connections, the model rates a person's brain as older than what is typical for their age peers.
Across all three cohorts, a higher BAI was associated with worse performance on a subset of tasks that require holding information in mind and planning actions, and with more pronounced symptoms of depression. The specific task batteries differed between cohorts, but the direction of the association was consistent.
In 182 participants, MRI, bacterial DNA analysis, and stool metabolite measurements were all available. When tested on data the model had not seen during tuning, the combined microbiota and metabolite profile captured 2.1% of BAI variance.
The authors propose repeating MRI, cognitive tests, mood scales, and stool analysis in the same individuals over a period of years. Such longitudinal measurements would show how these indicators change together over time.