MRI may predict decline in early-onset Alzheimer’s

MRI scans could help predict how quickly people with early-onset Alzheimer’s disease progress from mild cognitive impairment to dementia. Researchers at the Mass General Brigham Neuroscience Institute found that greater atrophy in specific brain regions was associated with faster progression. The findings could eventually help patients, families and clinicians plan future care more effectively.

Early-onset Alzheimer’s disease develops before the age of 65. Although it progresses faster on average than Alzheimer’s that develops later in life, the rate of decline varies considerably between individuals. This makes it difficult to predict when cognitive problems will begin to significantly affect daily functioning.

Brain atrophy linked to faster progression

The researchers analyzed data from 130 people aged 40 to 64 with mild cognitive impairment caused by early-onset Alzheimer’s disease. They also included 97 cognitively healthy people of similar ages. Participants came from the Longitudinal Early-Onset Alzheimer’s Disease Study (LEADS), a US multicenter research consortium. During an average follow-up period of approximately two years, 84 participants with early-onset Alzheimer’s, around 65 percent, progressed from mild cognitive impairment to dementia.

MRI scans taken at the beginning of the study revealed a relationship between brain atrophy and the speed of progression. Participants with greater shrinkage in regions involved in memory and thinking were more likely to develop dementia sooner. The risk of progression increased as the degree of atrophy rose.

According to co-senior author Alexandra Touroutoglou, an MRI-based prognosis could be particularly useful because MRI is already part of the diagnostic process. Better estimates of disease progression could help families make decisions about future care and living arrangements, while supporting clinicians in personalizing care.

Improving prediction

The researchers incorporated the MRI measurement into a model that already considered age, sex and performance on a baseline cognitive test. Adding brain atrophy improved the model’s ability to predict the risk of progression to dementia. Co-senior author Mark Eldaief said quantitative measurements of atrophy in vulnerable brain regions could eventually provide information about the timing of functional decline that clinicians currently cannot reliably offer patients.

The researchers caution, however, that the model was developed and tested predominantly in people with memory-related mild cognitive impairment. It may therefore be less applicable to forms of early-onset Alzheimer’s that initially affect behavior, language, vision or spatial abilities.

Additional research is needed before the MRI measure can be used for individual prognosis in clinical practice. The researchers want to validate the approach in more diverse populations and determine whether it can predict outcomes such as institutionalization or mortality. Future studies could also combine MRI findings with biomarkers for amyloid and tau. Such an approach may ultimately help identify patients who require closer monitoring or earlier care planning as their disease progresses.

AI model

Earlier this year, researchers at the University of California San Francisco developed an AI model that can predict the progression of Alzheimer’s disease using a single baseline MRI scan and basic demographic data. The approach could reduce dependence on cognitive testing, PET scans, genetic analyses and fluid biomarkers.

The multitask deep learning model simultaneously analyzes brain tissue and predicts Alzheimer’s diagnosis as well as current and future cognitive scores. By distinguishing grey matter, white matter and cerebrospinal fluid, the system gains a more detailed understanding of brain structure. In tests, it outperformed existing AI methods.

The researchers trained and validated the model using several datasets, including scans from healthy adults. This helped distinguish normal ageing from disease-related changes and improved generalisability. According to the researchers, the method could make prediction of cognitive decline faster and more accessible, particularly in settings where specialist testing and expertise are limited.