In Alzheimer's, two small brain regions shrink faster than the hippocampus
In brain scans of about 1,150 older adults from two US studies, the amygdala and the entorhinal cortex lost volume faster than the hippocampus, the region doctors usually look at, and the loss tracked where the protein tau had built up. Adding these regions to a computer test improved the detection of very early memory problems only modestly.
When doctors look at a brain scan for early signs of Alzheimer's disease, they usually check the hippocampus, the seahorse-shaped memory centre deep in each side of the brain. A new analysis of two US studies, one of them run at Johns Hopkins, argues that two neighbouring regions deserve as much attention: the amygdala, an almond-shaped cluster of nerve cells best known for handling emotions, and the entorhinal cortex, a thin strip of cortex that acts as the main gateway between the hippocampus and the rest of the brain. In two groups of older adults followed with repeated scans, both regions shrank faster than the hippocampus at every stage of the disease.
The result fits with what pathologists have long seen under the microscope. In Alzheimer's, tangles of the protein tau first appear in and around the entorhinal cortex and only later reach the hippocampus. The amygdala has received less attention, but tau-imaging studies have found tangles there early, too. Brain shrinkage on MRI is thought to reflect nerve cells and their connections being lost where these tangles form.
What they did
The researchers used data from two long-running US studies. The Alzheimer's Disease Neuroimaging Initiative (ADNI) contributed 1,019 people: 387 with normal thinking or only self-reported memory worries, 523 with mild cognitive impairment (memory or thinking problems that are noticeable but don't yet disrupt daily life) and 109 with Alzheimer's dementia. The BIOCARD study, run at the Johns Hopkins School of Medicine, added 130 people. Everyone had at least three MRI scans over time, so the team could measure how fast each region lost volume, expressed as percent per year. Average ages in the groups were between about 71 and 75.
A deep-learning program, trained on 665 hand-drawn outlines, marked out the regions on each scan. Some participants also had PET scans, which use a weakly radioactive tracer to show where tau and amyloid, the two hallmark proteins of Alzheimer's, have built up.
What they found
- The same order in both studies. The entorhinal region and the amygdala shrank at roughly similar rates, both clearly faster than the hippocampus. In ADNI, among people with Alzheimer's dementia, the entorhinal region lost 8.40% of its volume per year, the amygdala 7.19% and the hippocampus 3.23%. In people with mild cognitive impairment the figures were 3.53%, 3.26% and 1.65%. Even in the group with normal thinking, the hippocampus shrank more slowly (0.83% a year) than the other two (1.84% and 1.69%). BIOCARD showed the same order.
- Shrinking means thinning. In the entorhinal cortex, volume loss went hand in hand with the cortex becoming thinner, which the team sees as a sign that the measurement reflects real loss of tissue.
- Not all of the amygdala is hit equally. In a subgroup of ADNI participants with mild cognitive impairment or dementia, the inner (medial) parts of the amygdala shrank by about 6 to 8% a year, the outer (lateral) part by 2.52%. That matches where tau tangles are densest in brain tissue examined after death.
- Linked to tau. People with more tau on PET in a region tended to lose volume faster in that same region.
- A small gain for early detection. A computer model told apart people with normal thinking from those with early mild cognitive impairment slightly better when it used the amygdala and entorhinal measures than when it used the hippocampus alone. Its score for accuracy (the area under the curve, where 0.5 is a coin toss and 1 is perfect) rose from 0.635 to 0.665. Most of the gain came from fewer healthy people being wrongly flagged.
What to keep in mind
- Detection is still modest. The authors call the classification performance "moderate". An accuracy score of 0.665 is far from a test that could tell an individual whether they are developing Alzheimer's.
- Associations, not proof of sequence in each person. The study compares groups at different stages and links shrinkage to tau; it does not show that every person's amygdala shrinks first. News reports said the changes may begin up to a decade before memory loss. That figure comes from an earlier analysis the authors cite, not from the measurements in this paper.
- The small regions are harder to measure. The deep-learning outlines matched hand-drawn ones less closely for the amygdala and entorhinal cortex than for the hippocampus, especially in ADNI, where image quality varied more.
- One cohort is small. BIOCARD had 130 participants, and the amygdala subregion analysis was done in a subgroup of ADNI.
- The paper has no separate limitations section, so most of these points are ours, drawn from its results. One author declares equity in a company, AnatomyWorks.
Why it matters
The authors point out that treatments may work best in the earliest stages of Alzheimer's, which is why there is so much interest in spotting the disease before memory clearly fails. MRI is already a routine scan. This study suggests that MRI readers looking only at the hippocampus may be watching a region that changes later than its neighbours, and that the amygdala and entorhinal cortex carry extra information. The gain is small for now, and the measure would have to be tested in new groups of people before anyone could use it in a clinic. The team has made its analysis tools publicly available for others to try.