Brain Age Gap on MRI a Novel, Reliable Marker for Dementia Risk?

Brain Age Gap a Novel, Reliable Marker for Dementia Risk?

Megan Brooks

October 04, 2019

The gap between an individual’s brain age, as predicted on the basis of MRI, and their chronologic age may serve as a biomarker for gauging dementia risk, new research suggests.

“While we don’t have the treatment yet for dementia, it is important to diagnose it as early as possible,” Gennady Roshchupkin, PhD, Erasmus MC University Medical Center, Rotterdam, the Netherlands, told Medscape Medical News.

“For example, the person with large age gap can be invited for next MRI in a few years to evaluate if the gap remained the same or the risk now increased. Such imaging biomarker can be used together with others or in combination with genetic tests,” said Roshchupkin.

The study was published online October 1 in Proceedings of the National Academy of Sciences.

Machine Learning

The researchers built a convolutional neural network to predict brain aging. They employed deep learning to train it using brain scans of 3688 dementia-free individuals (mean age, 66 years; 55% women) in the Rotterdam Study. At least 5 years’ worth of follow-up clinical information of the participants was available.

In Cox proportional hazard models, the gap between predicted brain age (determined on the basis of the density of the brain’s gray matter) and chronologic age was significantly associated with incident dementia (hazard ratio, 1.11; 95% confidence interval, 1.06 – 1.15). The association remained significant after adjusting for hippocampal volume.

“The age gap has the potential to be utilized alongside other clinical risk factors and biomarkers to separate the population into categories with sufficiently distinct degrees of risk to drive clinical or personal decision-making, eg, dementia screening and informed life planning,” the investigators write.

“Before implemented in clinical practice, additional experiments should be done,” said Roshchupkin.

“Deep learning models are known to be biased in some situations towards gender, ethnicity, etc. We should be careful when generalizing models trained on a specific population. Therefore, for example, training such model on more diverse, multiethnic population can be a next step towards clinical usage,” said Roshchupkin.

Landmark Study 

Reached for comment, Cyrus Raji, MD, PhD, assistant professor of neuroradiology at Washington University in St. Louis, Missouri, said this is a “landmark paper showing that MRI scans can be used to determine brain age ― the actual age of a person’s brain compared to their chronological age.

“Moreover, they have demonstrated that risk of dementia such as Alzheimer’s disease is higher in persons with an older brain on their MRI scan compared to their chronological age. This work can help determine who is at risk for dementia so that better treatments and preventive measure scan be successfully applied,” said Raji.

Rebecca Edelmayer, PhD, director, scientific engagement for the Alzheimer’s Association, told Medscape Medical News, “Machine learning and other aspects of artificial intelligence to make medical predictions is an exciting area of research that is still evolving.

“These dementia-related data are interesting and important to publish, particularly because they involve use of imaging measurements made across time. However, the work described in this new publication is very preliminary,” said Edelmayer.

“The population involved in this study is not diverse, the results of this study are not generalizable (as the authors point out), and the technology is not yet practical for clinical use. At this early point, further research is necessary with additional, more diverse training datasets to develop the accuracy needed for predictive utility,” she added.

Edelmayer also noted that the Alzheimer’s Association has funded several studies in recent years that ttested the ability of “cutting-edge computer science and statistical techniques” to predict progression to dementia in people with mild cognitive impairment, identify individuals at high risk of dementia using big data sets from population-based surveys, and predict an individual with dementia’s likelihood of falling by identifying subtle changes in a person’s gait.

The Rotterdam Study is funded by the Erasmus Medical Center and Erasmus University, Rotterdam, Netherlands Organization for the Health Research and Development, the Research Institute for Diseases in the Elderly, the Ministry of Education, Culture and Science, the Ministry for Health, Welfare and Sports, the European Commission (Directorate-General XII), and the Municipality of Rotterdam. The authors, Raji, and Edelmayer have disclosed no relevant financial relationships. 

Proc Natl Acad Sci. Published online October 1, 2019. Abstract