Benchmarking the generalizability of brain age models: Challenges posed by scanner variance and prediction bias

Mega-analysis of the brain-age gap in substance use disorder: An ENIGMA Addiction working group study

BACKGROUND AND AIMS

The brain age gap (BAG), calculated as the difference between a machine learning model-based predicted brain age and chronological age, has been increasingly investigated in psychiatric disorders. Tobacco and alcohol use are associated with increased BAG; however, no studies have compared global and regional BAG across substances other than alcohol and tobacco. This study aimed to compare global and regional estimates of brain age in individuals with substance use disorders and healthy controls.

DESIGN

This was a cross-sectional study.

SETTING

This is an Enhancing Neuro Imaging through Meta-Analysis Consortium (ENIGMA) Addiction Working Group study including data from 38 global sites.

PARTICIPANTS

This study included 2606 participants, of whom 1725 were cases with a substance use disorder and 881 healthy controls.

MEASUREMENTS

This study used the Kaufmann brain age prediction algorithms to generate global and regional brain age estimates using T1 weighted magnetic resonance imaging (MRI) scans. We used linear mixed effects models to compare global and regional (FreeSurfer lobestrict output) BAG (i.e. predicted minus chronological age) between individuals with one of five primary substance use disorders as well as healthy controls.

FINDINGS

Alcohol use disorder (β = -5.49, t = -5.51, p < 0.001) was associated with higher global BAG, whereas amphetamine-type stimulant use disorder (β = 3.44, t = 2.42, p = 0.02) was associated with lower global BAG in the separate substance-specific models.

CONCLUSIONS

People with alcohol use disorder appear to have a higher brain-age gap than people without alcohol use disorder, which is consistent with other evidence of the negative impact of alcohol on the brain.

A review of artificial intelligence-based brain age estimation and its applications for related diseases

The study of brain age has emerged over the past decade, aiming to estimate a person's age based on brain imaging scans. Ideally, predicted brain age should match chronological age in healthy individuals. However, brain structure and function change in the presence of brain-related diseases. Consequently, brain age also changes in affected individuals, making the brain age gap (BAG)-the difference between brain age and chronological age-a potential biomarker for brain health, early screening, and identifying age-related cognitive decline and disorders. With the recent successes of artificial intelligence in healthcare, it is essential to track the latest advancements and highlight promising directions. This review paper presents recent machine learning techniques used in brain age estimation (BAE) studies. Typically, BAE models involve developing a machine learning regression model to capture age-related variations in brain structure from imaging scans of healthy individuals and automatically predict brain age for new subjects. The process also involves estimating BAG as a measure of brain health. While we discuss recent clinical applications of BAE methods, we also review studies of biological age that can be integrated into BAE research. Finally, we point out the current limitations of BAE's studies.

Exploring associations between psychotic experiences and structural brain age: a population-based study in late adolescence

Neuroimaging studies show advanced structural "brain age" in schizophrenia and related psychotic disorders, potentially reflecting aberrant brain ageing or maturation. The extent to which altered brain age is associated with subthreshold psychotic experiences (PE) in youth remains unclear. We investigated the association between PE and brain-predicted age difference (brain-PAD) in late adolescence using a population-based sample of 117 participants with PE and 115 without PE (aged 19-21 years) from the Avon Longitudinal Study of Parents and Children. Brain-PAD was estimated using a publicly available machine learning model previously trained on a combination of region-wise T1-weighted grey-matter measures. We found little evidence for an association between PEs and brain-PAD after adjusting for age and sex (Cohen's d = -0.21 [95% CI -0.47, 0.05], p = 0.11). While there was some evidence for lower brain-PAD in those with PEs relative to those without PEs after additionally adjusting for parental social class (Cohen's d = -0.31 [95% CI -0.58, -0.03], p = 0.031) or birth weight (Cohen's d = -0.29 [95% CI -0.55, -0.03], p = 0.038), adjusting for maternal education or childhood IQ did not alter the primary results. These findings do not support the notion of advanced brain age in older adolescents with PEs. However, they weakly suggest there might be a younger-looking brain in those individuals, indicative of subtle delays in structural brain maturation. Future studies with larger samples covering a wider age range and multimodal measures could further investigate brain age as a marker of psychotic experiences in youth.

A perspective on brain-age estimation and its clinical promise

Brain-age estimation has gained increased attention in the neuroscientific community owing to its potential use as a biomarker of brain health. The difference between estimated and chronological age based on neuroimaging data enables a unique perspective on brain development and aging, with multiple open questions still remaining in the brain-age research field. This Perspective presents an overview of current advancements in the field and envisions the future evolution of the brain-age framework before its potential deployment in hospital settings.

Machine Learning and Deep Learning Approaches in Lifespan Brain Age Prediction: A Comprehensive Review

The concept of 'brain age', derived from neuroimaging data, serves as a crucial biomarker reflecting cognitive vitality and neurodegenerative trajectories. In the past decade, machine learning (ML) and deep learning (DL) integration has transformed the field, providing advanced models for brain age estimation. However, achieving precise brain age prediction across all ages remains a significant analytical challenge. This comprehensive review scrutinizes advancements in ML- and DL-based brain age prediction, analyzing 52 peer-reviewed studies from 2020 to 2024. It assesses various model architectures, highlighting their effectiveness and nuances in lifespan brain age studies. By comparing ML and DL, strengths in forecasting and methodological limitations are revealed. Finally, key findings from the reviewed articles are summarized and a number of major issues related to ML/DL-based lifespan brain age prediction are discussed. Through this study, we aim at the synthesis of the current state of brain age prediction, emphasizing both advancements and persistent challenges, guiding future research, technological advancements, and improving early intervention strategies for neurodegenerative diseases.

Resting-state functional connectivity correlates of brain structural aging in schizophrenia

A large body of research has shown that schizophrenia patients demonstrate increased brain structural aging. Although this process may be coupled with aberrant changes in intrinsic functional architecture of the brain, they remain understudied. We hypothesized that there are brain regions whose whole-brain functional connectivity at rest is differently associated with brain structural aging in schizophrenia patients compared to healthy controls. Eighty-four male schizophrenia patients and eighty-six male healthy controls underwent structural MRI and resting-state fMRI. The brain-predicted age difference (b-PAD) was a measure of brain structural aging. Resting-state fMRI was applied to obtain global correlation (GCOR) maps comprising voxelwise values of the strength and sign of functional connectivity of a given voxel with the rest of the brain. Schizophrenia patients had higher b-PAD compared to controls (mean between-group difference + 2.9 years). Greater b-PAD in schizophrenia patients, compared to controls, was associated with lower whole-brain functional connectivity of a region in frontal orbital cortex, inferior frontal gyrus, Heschl's Gyrus, plana temporale and polare, insula, and opercular cortices of the right hemisphere (rFTI). According to post hoc seed-based correlation analysis, decrease of functional connectivity with the posterior cingulate gyrus, left superior temporal cortices, as well as right angular gyrus/superior lateral occipital cortex has mainly driven the results. Lower functional connectivity of the rFTI was related to worse verbal working memory and language production. Our findings demonstrate that well-established frontotemporal functional abnormalities in schizophrenia are related to increased brain structural aging.

Brain age has limited utility as a biomarker for capturing fluid cognition in older individuals

One well-known biomarker candidate that supposedly helps capture fluid cognition is Brain Age, or a predicted value based on machine-learning models built to predict chronological age from brain MRI. To formally evaluate the utility of Brain Age for capturing fluid cognition, we built 26 age-prediction models for Brain Age based on different combinations of MRI modalities, using the Human Connectome Project in Aging (n=504, 36-100 years old). First, based on commonality analyses, we found a large overlap between Brain Age and chronological age: Brain Age could uniquely add only around 1.6% in explaining variation in fluid cognition over and above chronological age. Second, the age-prediction models that performed better at predicting chronological age did NOT necessarily create better Brain Age for capturing fluid cognition over and above chronological age. Instead, better-performing age-prediction models created Brain Age that overlapped larger with chronological age, up to around 29% out of 32%, in explaining fluid cognition. Third, Brain Age missed around 11% of the total variation in fluid cognition that could have been explained by the brain variation. That is, directly predicting fluid cognition from brain MRI data (instead of relying on Brain Age and chronological age) could lead to around a 1/3-time improvement of the total variation explained. Accordingly, we demonstrated the limited utility of Brain Age as a biomarker for fluid cognition and made some suggestions to ensure the utility of Brain Age in explaining fluid cognition and other phenotypes of interest.

Brain asymmetries from mid- to late life and hemispheric brain age

The human brain demonstrates structural and functional asymmetries which have implications for ageing and mental and neurological disease development. We used a set of magnetic resonance imaging (MRI) metrics derived from structural and diffusion MRI data in N=48,040 UK Biobank participants to evaluate age-related differences in brain asymmetry. Most regional grey and white matter metrics presented asymmetry, which were higher later in life. Informed by these results, we conducted hemispheric brain age (HBA) predictions from left/right multimodal MRI metrics. HBA was concordant to conventional brain age predictions, using metrics from both hemispheres, but offers a supplemental general marker of brain asymmetry when setting left/right HBA into relationship with each other. In contrast to WM brain asymmetries, left/right discrepancies in HBA are lower at higher ages. Our findings outline various sex-specific differences, particularly important for brain age estimates, and the value of further investigating the role of brain asymmetries in brain ageing and disease development.

Prediction of brain age using structural magnetic resonance imaging: A comparison of accuracy and test-retest reliability of publicly available software packages

Brain age prediction algorithms using structural magnetic resonance imaging (MRI) aim to assess the biological age of the human brain. The difference between a person's chronological age and the estimated brain age is thought to reflect deviations from a normal aging trajectory, indicating a slower or accelerated biological aging process. Several pre-trained software packages for predicting brain age are publicly available. In this study, we perform a comparison of such packages with respect to (1) predictive accuracy, (2) test-retest reliability, and (3) the ability to track age progression over time. We evaluated the six brain age prediction packages: brainageR, DeepBrainNet, brainage, ENIGMA, pyment, and mccqrnn. The accuracy and test-retest reliability were assessed on MRI data from 372 healthy people aged between 18.4 and 86.2 years (mean 38.7 ± 17.5 years). All packages showed significant correlations between predicted brain age and chronological age (r = 0.66-0.97, p < 0.001), with pyment displaying the strongest correlation. The mean absolute error was between 3.56 (pyment) and 9.54 years (ENIGMA). brainageR, pyment, and mccqrnn were superior in terms of reliability (ICC values between 0.94-0.98), as well as predicting age progression over a longer time span. Of the six packages, pyment and brainageR consistently showed the highest accuracy and test-retest reliability.

Considerations on brain age predictions from repeatedly sampled data across time

INTRODUCTION

Brain age, the estimation of a person's age from magnetic resonance imaging (MRI) parameters, has been used as a general indicator of health. The marker requires however further validation for application in clinical contexts. Here, we show how brain age predictions perform for the same individual at various time points and validate our findings with age-matched healthy controls.

METHODS

We used densely sampled T1-weighted MRI data from four individuals (from two densely sampled datasets) to observe how brain age corresponds to age and is influenced by acquisition and quality parameters. For validation, we used two cross-sectional datasets. Brain age was predicted by a pretrained deep learning model.

RESULTS

We found small within-subject correlations between age and brain age. We also found evidence for the influence of field strength on brain age which replicated in the cross-sectional validation data and inconclusive effects of scan quality.

CONCLUSION

The absence of maturation effects for the age range in the presented sample, brain age model bias (including training age distribution and field strength), and model error are potential reasons for small relationships between age and brain age in densely sampled longitudinal data. Clinical applications of brain age models should consider of the possibility of apparent biases caused by variation in the data acquisition process.

Bio-psycho-social factors' associations with brain age: a large-scale UK Biobank diffusion study of 35,749 participants

Brain age refers to age predicted by brain features. Brain age has previously been associated with various health and disease outcomes and suggested as a potential biomarker of general health. Few previous studies have systematically assessed brain age variability derived from single and multi-shell diffusion magnetic resonance imaging data. Here, we present multivariate models of brain age derived from various diffusion approaches and how they relate to bio-psycho-social variables within the domains of sociodemographic, cognitive, life-satisfaction, as well as health and lifestyle factors in midlife to old age (N = 35,749, 44.6-82.8 years of age). Bio-psycho-social factors could uniquely explain a small proportion of the brain age variance, in a similar pattern across diffusion approaches: cognitive scores, life satisfaction, health and lifestyle factors adding to the variance explained, but not socio-demographics. Consistent brain age associations across models were found for waist-to-hip ratio, diabetes, hypertension, smoking, matrix puzzles solving, and job and health satisfaction and perception. Furthermore, we found large variability in sex and ethnicity group differences in brain age. Our results show that brain age cannot be sufficiently explained by bio-psycho-social variables alone. However, the observed associations suggest to adjust for sex, ethnicity, cognitive factors, as well as health and lifestyle factors, and to observe bio-psycho-social factor interactions' influence on brain age in future studies.

A systematic review of multimodal brain age studies: Uncovering a divergence between model accuracy and utility

Brain aging is a complex, multifaceted process that can be challenging to model in ways that are accurate and clinically useful. One of the most common approaches has been to apply machine learning to neuroimaging data with the goal of predicting age in a data-driven manner. Building on initial brain age studies that were derived solely from T1-weighted scans (i.e., unimodal), recent studies have incorporated features across multiple imaging modalities (i.e., "multimodal"). In this systematic review, we show that unimodal and multimodal models have distinct advantages. Multimodal models are the most accurate and sensitive to differences in chronic brain disorders. In contrast, unimodal models from functional magnetic resonance imaging were most sensitive to differences across a broad array of phenotypes. Altogether, multimodal imaging has provided us valuable insight for improving the accuracy of brain age models, but there is still much untapped potential with regard to achieving widespread clinical utility.

The effect of weight loss following 18 months of lifestyle intervention on brain age assessed with resting-state functional connectivity

Background

Obesity negatively impacts multiple bodily systems, including the central nervous system. Retrospective studies that estimated chronological age from neuroimaging have found accelerated brain aging in obesity, but it is unclear how this estimation would be affected by weight loss following a lifestyle intervention.

Methods

In a sub-study of 102 participants of the Dietary Intervention Randomized Controlled Trial Polyphenols Unprocessed Study (DIRECT-PLUS) trial, we tested the effect of weight loss following 18 months of lifestyle intervention on predicted brain age based on magnetic resonance imaging (MRI)-assessed resting-state functional connectivity (RSFC). We further examined how dynamics in multiple health factors, including anthropometric measurements, blood biomarkers, and fat deposition, can account for changes in brain age.

Results

To establish our method, we first demonstrated that our model could successfully predict chronological age from RSFC in three cohorts (n=291;358;102). We then found that among the DIRECT-PLUS participants, 1% of body weight loss resulted in an 8.9 months' attenuation of brain age. Attenuation of brain age was significantly associated with improved liver biomarkers, decreased liver fat, and visceral and deep subcutaneous adipose tissues after 18 months of intervention. Finally, we showed that lower consumption of processed food, sweets and beverages were associated with attenuated brain age.

Conclusions

Successful weight loss following lifestyle intervention might have a beneficial effect on the trajectory of brain aging.

Funding

The German Research Foundation (DFG), German Research Foundation - project number 209933838 - SFB 1052; B11, Israel Ministry of Health grant 87472511 (to I Shai); Israel Ministry of Science and Technology grant 3-13604 (to I Shai); and the California Walnuts Commission 09933838 SFB 105 (to I Shai).