Grey Matter Age Prediction as a Biomarker for Risk of Dementia: A Population-based Study

Johnny Wang, Maria J. Knol, Aleksei Tiulpin, Florian Dubost, Marleen de Bruijne, Meike W. Vernooij, Hieab H.h. Adams, M. Arfan Ikram, Wiro J. Niessen, Gennady V. RoshchupkinbioRxiv 2019 · DOI

TL;DR

This study uses deep learning to predict "brain age" from MRI scans and finds that a larger gap between predicted brain age and chronological age is a significant predictor of future dementia risk. This "brain age gap" could serve as a valuable, non-invasive biomarker for early dementia screening.

What Problem Does This Paper Address?

As we age, our brains naturally change. However, neurodegenerative diseases like dementia also impact brain structure and function, often years before clinical symptoms appear. Identifying individuals at risk early on is crucial for timely intervention and management. While chronological age is a known risk factor, researchers are exploring more sensitive biomarkers. One promising avenue is "brain age," a prediction of a person's brain age based on imaging data. The gap between this predicted brain age and a person's actual chronological age (the "brain age gap") has been hypothesized to reflect underlying brain health and potentially predict future cognitive decline. However, previous studies were often cross-sectional, making it difficult to establish a causal link between the age gap and the development of dementia over time. This paper aims to bridge that gap by using a longitudinal dataset to investigate if the brain age gap is a reliable predictor of incident dementia.

Methods

The researchers employed a Convolutional Neural Network (CNN), a type of deep learning model particularly well-suited for image analysis, to predict brain age. They used MRI-derived grey matter (GM) density maps as input. The CNN was trained on a large dataset and its performance was evaluated using Mean Absolute Error (MAE) for age prediction accuracy and Intraclass Correlation Coefficient (ICC) for reproducibility. To understand which brain regions were most influential in the age prediction, they utilized Gradient-weighted Class Activation Mapping (Grad-CAM) to generate "attention maps." Finally, logistic regression and Cox proportional hazards models were used to assess the association between the brain age gap and the incidence of dementia, controlling for various demographic and clinical factors.

Datasets

The study utilized data from the Rotterdam Study, a large, prospective, population-based cohort study in the Netherlands.

  • Training, validation, and testing sets were derived from participants who did not have dementia or stroke at baseline.
  • A separate independent dataset was used for subjects who developed dementia during the follow-up period.
  • A total of 5496 participants were included in the analysis.

Key Insights and Ideas

  • Brain Age as a Window into Brain Health: The study reinforces the idea that our brains don't age uniformly. The discrepancy between our biological brain age and our chronological age can reveal underlying health status.
  • The Amygdala and Hippocampus are Key Players: The attention maps pinpointed the amygdala and hippocampus as particularly important regions for predicting brain age. These areas are known to be involved in memory and emotion, and are often affected early in neurodegenerative processes.
  • Longitudinal Data is Crucial: Moving beyond cross-sectional studies, this research provides strong evidence for the predictive power of the brain age gap in a real-world, longitudinal setting, showing it's not just a snapshot but a predictor of future events.
  • A Complementary Biomarker: The brain age gap isn't meant to replace existing dementia risk factors but to complement them, offering a new layer of information for risk stratification.

Key Results

  • Accurate Brain Age Prediction: The CNN achieved a Mean Absolute Error (MAE) of 4.45±3.59 years for brain age prediction, indicating a good level of accuracy. Reproducibility was high, with an ICC of 0.97.
  • Significant Dementia Risk Association: The brain age gap was significantly associated with an increased risk of incident dementia. For every year the predicted brain age exceeded chronological age, the odds ratio for dementia increased by 1.11 (95% CI 1.05-1.16), and the hazard ratio by 1.11 (95% CI 1.06-1.15), even after adjusting for multiple confounding factors.
  • Attention Maps Highlighted Key Regions: The Grad-CAM analysis revealed that the nucleus accumbens (0.89) and amygdala (0.71) had the highest mean attention map intensities, suggesting their strong contribution to age prediction.

Limitations and Future Work

  • Perfect Prediction Not Achieved: The authors acknowledge that they couldn't perfectly predict the age of healthy subjects solely from MRI, attributing this to the inherent biological similarity of brains within a certain age range.
  • Exclusion of Existing Conditions: The model was trained excluding individuals with pre-existing dementia or stroke. While this helps isolate the predictive power for incident dementia, it means the model's direct applicability to these populations needs further investigation.
  • Generalizability: The current CNN model might struggle with unfamiliar datasets, limiting its generalizability. The authors suggest training models on more diverse datasets to improve this.
  • Further Research: Future work could explore the specific gap-associated features that influence these differences and investigate how to integrate this information into models to improve accessibility and accuracy. Further research into the association between the age gap and dementia, compared to other biomarkers, is also warranted.

Summary updated

Abstract

Key Points Question Is the gap between brain age predicted from MRI and chronological age associated with incident dementia in a general population of Dutch adults? Findings Brain age was predicted using a deep learning model, using MRI-derived grey matter density maps. In a population based study including 5496 participants, the observed gap was significantly associated with the risk of dementia. Meaning The gap between MRI-brain predicted and chronological age is potentially a biomarker for dementia risk screening. Abstract Importance The gap between predicted brain age using magnetic resonance imaging (MRI) and chronological age may serve as biomarker for early-stage neurodegeneration and potentially as a risk indicator for dementia. However, owing to the lack of large longitudinal studies, it has been challenging to validate this link. Objective We aimed to investigate the utility of such a gap as a risk biomarker for incident dementia in a general Dutch population, using a deep learning approach for predicting brain age based on MRI-derived grey matter maps. Design Data was collected from participants of the cohort-based Rotterdam Study who underwent brain magnetic resonance imaging between 2006 and 2015. This study was performed in a longitudinal setting and all participant were followed up for incident dementia until 2016. Setting The Rotterdam Study is a prospective population-based study, initiated in 1990 in the suburb Ommoord of in Rotterdam, the Netherlands. Participants At baseline, 5496 dementia- and stroke-free participants (mean age 64.67±9.82, 54.73% women) were scanned and screened for incident dementia. During 6.66±2.46 years of follow-up, 159 people developed dementia. Main outcomes and measures We built a convolutional neural network (CNN) model to predict brain age based on its MRI. Model prediction performance was measured in mean absolute error (MAE). Reproducibility of prediction was tested using the intraclass correlation coefficient (ICC) computed on a subset of 80 subjects. Logistic regressions and Cox proportional hazards were used to assess the association of the age gap with incident dementia, adjusted for years of education, ApoEε4 allele carriership, grey matter volume and intracranial volume. Additionally, we computed the attention maps of CNN, which shows which brain regions are important for age prediction. Results MAE of brain age prediction was 4.45±3.59 years and ICC was 0.97 (95% confidence interval CI=0.96-0.98). Logistic regression and Cox proportional hazards models showed that the age gap was significantly related to incident dementia (odds ratio OR=1.11 and 95% confidence intervals CI=1.05-1.16; hazard ratio HR=1.11 and 95% CI=1.06-1.15, respectively). Attention maps indicated that grey matter density around the amygdalae and hippocampi primarily drive the age estimation. Conclusion and relevance We show that the gap between predicted and chronological brain age is a biomarker associated with risk of dementia development. This suggests that it can be used as a biomarker, complimentary to those that are known, for dementia risk screening.
BibTeX
@misc{wang2019grey,
  author = {Johnny Wang and Maria J. Knol and Aleksei Tiulpin and Florian Dubost and Marleen de Bruijne and Meike W. Vernooij and Hieab H.H. Adams and M. Arfan Ikram and Wiro J. Niessen and Gennady V. Roshchupkin},
  title = {{Grey Matter Age Prediction as a Biomarker for Risk of Dementia: A Population-based Study}},
  year = {2019},
  doi = {10.1101/518506},
  howpublished = {bioRxiv},
}
Figure