Anatomically interpretable deep learning of brain age captures domain-specific cognitive impairment

ds-FCRN: three-dimensional dual-stream fully convolutional residual networks and transformer-based global-local feature learning for brain age prediction

The brain undergoes atrophy and cognitive decline with advancing age. The utilization of brain age prediction represents a pioneering methodology in the examination of brain aging. This study aims to develop a deep learning model with high predictive accuracy and interpretability for brain age prediction tasks. The gray matter (GM) density maps obtained from T1 MRI data of 16,377 healthy participants aged 45 to 82 years from the UKB database were included in this study (mean age, 64.27 ± 7.52 , 7811 men). We propose an innovative deep learning architecture for predicting brain age based on GM density maps. The architecture combines a 3D dual-stream fully convolutional residual network (ds-FCRN) with a Transformer-based global-local feature learning paradigm to enhance prediction accuracy. Moreover, we employed Shapley values to elucidate the influence of various brain regions on prediction precision. On a test set of 3,276 healthy subjects (mean age, 64.15 ± 7.45 , 1561 men), our 3D ds-FCRN model achieved a mean absolute error of 2.2 years in brain age prediction, outperforming existing models on the same dataset. The posterior interpretation revealed that the temporal lobe plays the most significant role in the brain age prediction process, while frontal lobe aging is associated with the greatest number of lifestyle factors. Our designed 3D ds-FCRN model achieved high predictive accuracy and high decision transparency. The brain age vectors constructed using Shapley values provided brain region-level insights into life factors associated with abnormal brain aging.

Deep learning and generative artificial intelligence in aging research and healthy longevity medicine

With the global population aging at an unprecedented rate, there is a need to extend healthy productive life span. This review examines how Deep Learning (DL) and Generative Artificial Intelligence (GenAI) are used in biomarker discovery, deep aging clock development, geroprotector identification and generation of dual-purpose therapeutics targeting aging and disease. The paper explores the emergence of multimodal, multitasking research systems highlighting promising future directions for GenAI in human and animal aging research, as well as clinical application in healthy longevity medicine.

Age-disproportionate atrophy in Alzheimer's disease and Parkinson's disease spectra

INTRODUCTION

Brain age gap (BAG), defined as the difference between MRI-predicted 'brain age' and chronological age, can capture information underlying various neurological disorders. We investigated the pathophysiological significance of the BAG across neurodegenerative disorders.

METHODS

We developed a brain age estimator using structural MRIs of healthy-aged individuals from one cohort study. Subsequently, we applied this estimator to people with Alzheimer's disease spectra (AD) and Parkinson's disease (PD) from another cohort study. We investigated brain sources responsible for BAGs among these groups.

RESULTS

Both AD and PD exhibited a positive BAG. Brain sources showed overlapping, yet partially segregated, neuromorphological differences between these groups. Furthermore, employing with t-distributed stochastic neighbor embedding on the brain sources, we subclassified PD into two groups with and without cognitive impairment.

DISCUSSION

Our findings suggest that brain age estimation becomes a clinically relevant method for finely stratifying neurodegenerative disorders.

Highlights

Brain age estimated from structure MRI data was greater than chronological age in patients with Alzheimer's disease/mild cognitive impairment or Parkinson's disease.Brain regions attributed to brain age estimation were located mainly in the fronto-temporo-parietal cortices but not in the motor cortex or subcortical regions.Brain sources responsible for the brain age gaps revealed roughly overlapping, yet partially segregated, neuromorphological differences between participants with Alzheimer's disease/mild cognitive impairment and Parkinson's disease.Participants with Parkinson's disease were subclassified into two groups (with and without cognitive impairment) based on brain sources responsible for the brain age gaps.

Disentangling Neurodegeneration From Aging in Multiple Sclerosis Using Deep Learning: The Brain-Predicted Disease Duration Gap

BACKGROUND AND OBJECTIVES

Disentangling brain aging from disease-related neurodegeneration in patients with multiple sclerosis (PwMS) is increasingly topical. The brain-age paradigm offers a window into this problem but may miss disease-specific effects. In this study, we investigated whether a disease-specific model might complement the brain-age gap (BAG) by capturing aspects unique to MS.

METHODS

In this retrospective study, we collected 3D T1-weighted brain MRI scans of PwMS to build (1) a cross-sectional multicentric cohort for age and disease duration (DD) modeling and (2) a longitudinal single-center cohort of patients with early MS as a clinical use case. We trained and evaluated a 3D DenseNet architecture to predict DD from minimally preprocessed images while age predictions were obtained with the DeepBrainNet model. The brain-predicted DD gap (the difference between predicted and actual duration) was proposed as a DD-adjusted global measure of MS-specific brain damage. Model predictions were scrutinized to assess the influence of lesions and brain volumes while the DD gap was biologically and clinically validated within a linear model framework assessing its relationship with BAG and physical disability measured with the Expanded Disability Status Scale (EDSS).

RESULTS

We gathered MRI scans of 4,392 PwMS (69.7% female, age: 42.8 ± 10.6 years, DD: 11.4 ± 9.3 years) from 15 centers while the early MS cohort included 749 sessions from 252 patients (64.7% female, age: 34.5 ± 8.3 years, DD: 0.7 ± 1.2 years). Our model predicted DD better than chance (mean absolute error = 5.63 years, R2 = 0.34) and was nearly orthogonal to the brain-age model (correlation between DD and BAGs: r = 0.06 [0.00-0.13], p = 0.07). Predictions were influenced by distributed variations in brain volume and, unlike brain-predicted age, were sensitive to MS lesions (difference between unfilled and filled scans: 0.55 years [0.51-0.59], p < 0.001). DD gap significantly explained EDSS changes (B = 0.060 [0.038-0.082], p < 0.001), adding to BAG (ΔR2 = 0.012, p < 0.001). Longitudinally, increasing DD gap was associated with greater annualized EDSS change (r = 0.50 [0.39-0.60], p < 0.001), with an incremental contribution in explaining disability worsening compared with changes in BAG alone (ΔR2 = 0.064, p < 0.001).

DISCUSSION

The brain-predicted DD gap is sensitive to MS-related lesions and brain atrophy, adds to the brain-age paradigm in explaining physical disability both cross-sectionally and longitudinally, and may be used as an MS-specific biomarker of disease severity and progression.

Brain age prediction and deviations from normative trajectories in the neonatal connectome

Structural and functional connectomes undergo rapid changes during the third trimester and the first month of postnatal life. Despite progress, our understanding of the developmental trajectories of the connectome in the perinatal period remains incomplete. Brain age prediction uses machine learning to estimate the brain's maturity relative to normative data. The difference between the individual's predicted and chronological age-or brain age gap (BAG)-represents the deviation from these normative trajectories. Here, we assess brain age prediction and BAGs using structural and functional connectomes for infants in the first month of life. We use resting-state fMRI and DTI data from 611 infants (174 preterm; 437 term) from the Developing Human Connectome Project (dHCP) and connectome-based predictive modeling to predict postmenstrual age (PMA). Structural and functional connectomes accurately predict PMA for term and preterm infants. Predicted ages from each modality are correlated. At the network level, nearly all canonical brain networks-even putatively later developing ones-generate accurate PMA prediction. Additionally, BAGs are associated with perinatal exposures and toddler behavioral outcomes. Overall, our results underscore the importance of normative modeling and deviations from these models during the perinatal period.

Complementary value of molecular, phenotypic, and functional aging biomarkers in dementia prediction

DNA methylation age (MA), brain age (BA), and frailty index (FI) are putative aging biomarkers linked to dementia risk. We investigated their relationship and combined potential for prediction of cognitive impairment and future dementia risk using the ADNI database. Of several MA algorithms, DunedinPACE and GrimAge2, associated with memory, were combined in a composite MA alongside BA and a data-driven FI in predictive analyses. Pairwise correlations between age- and sex-adjusted measures for MA (aMA), aBA, and aFI were low. FI outperformed BA and MA in all diagnostic tasks. A model including age, sex, and aFI achieved an area under the curve (AUC) of 0.94 for differentiating cognitively normal controls (CN) from dementia patients in a held-out test set. When combined with clinical biomarkers (apolipoprotein E ε4 allele count, memory, executive function), a model including aBA and aFI predicted 5-year dementia risk among MCI patients with an out-of-sample AUC of 0.88. In the prognostic model, BA and FI offered complementary value (both βs 0.50). The tested MAs did not improve predictions. Results were consistent across FI algorithms, with data-driven health deficit selection yielding the best performance. FI had a stronger adverse effect on prognosis in males, while BA's impact was greater in females. Our findings highlight the complementary value of BA and FI in dementia prediction. The results support a multidimensional view of dementia, including an intertwined relationship between the biomarkers, sex, and prognosis. The tested MA's limited contribution suggests caution in their use for individual risk assessment of dementia.

Anatomic Interpretability in Neuroimage Deep Learning: Saliency Approaches for Typical Aging and Traumatic Brain Injury

The black box nature of deep neural networks (DNNs) makes researchers and clinicians hesitant to rely on their findings. Saliency maps can enhance DNN explainability by suggesting the anatomic localization of relevant brain features. This study compares seven popular attribution-based saliency approaches to assign neuroanatomic interpretability to DNNs that estimate biological brain age (BA) from magnetic resonance imaging (MRI). Cognitively normal (CN) adults ( N = 13,394 , 5,900 males; mean age: 65.82 ± 8.89 years) are included for DNN training, testing, validation, and saliency map generation to estimate BA. To study saliency robustness to the presence of anatomic deviations from normality, saliency maps are also generated for adults with mild traumatic brain injury (mTBI, N = 214 , 135 males; mean age: 55.3 ± 9.9 years). We assess saliency methods' capacities to capture known anatomic features of brain aging and compare them to a surrogate ground truth whose anatomic saliency is known a priori. Anatomic aging features are identified most reliably by the integrated gradients method, which outperforms all others through its ability to localize relevant anatomic features. Gradient Shapley additive explanations, input × gradient, and masked gradient perform less consistently but still highlight ubiquitous neuroanatomic features of aging (ventricle dilation, hippocampal atrophy, sulcal widening). Saliency methods involving gradient saliency, guided backpropagation, and guided gradient-weight class attribution mapping localize saliency outside the brain, which is undesirable. Our research suggests the relative tradeoffs of saliency methods to interpret DNN findings during BA estimation in typical aging and after mTBI.

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.

Anatomic Interpretability in Neuroimage Deep Learning: Saliency Approaches for Typical Aging and Traumatic Brain Injury

The black box nature of deep neural networks (DNNs) makes researchers and clinicians hesitant to rely on their findings. Saliency maps can enhance DNN explainability by suggesting the anatomic localization of relevant brain features. This study compares seven popular attribution-based saliency approaches to assign neuroanatomic interpretability to DNNs that estimate biological brain age (BA) from magnetic resonance imaging (MRI). Cognitively normal (CN) adults (N = 13,394, 5,900 males; mean age: 65.82 ± 8.89 years) are included for DNN training, testing, validation, and saliency map generation to estimate BA. To study saliency robustness to the presence of anatomic deviations from normality, saliency maps are also generated for adults with mild traumatic brain injury (mTBI, N = 214, 135 males; mean age: 55.3 ± 9.9 years). We assess saliency methods' capacities to capture known anatomic features of brain aging and compare them to a surrogate ground truth whose anatomic saliency is known a priori. Anatomic aging features are identified most reliably by the integrated gradients method, which outperforms all others through its ability to localize relevant anatomic features. Gradient Shapley additive explanations, input × gradient, and masked gradient perform less consistently but still highlight ubiquitous neuroanatomic features of aging (ventricle dilation, hippocampal atrophy, sulcal widening). Saliency methods involving gradient saliency, guided backpropagation, and guided gradient-weight class attribution mapping localize saliency outside the brain, which is undesirable. Our research suggests the relative tradeoffs of saliency methods to interpret DNN findings during BA estimation in typical aging and after mTBI.

An estimate of the longitudinal pace of aging from a single brain scan predicts dementia conversion, morbidity, and mortality

To understand how aging affects functional decline and increases disease risk, it is necessary to develop accurate and reliable measures of how fast a person is aging. Epigenetic clocks measure aging but require DNA methylation data, which many studies lack. Using data from the Dunedin Study, we introduce an accurate and reliable measure for the rate of longitudinal aging derived from cross-sectional brain MRI: the Dunedin Pace of Aging Calculated from NeuroImaging or DunedinPACNI. Exporting this measure to the Alzheimer's Disease Neuroimaging Initiative and UK Biobank neuroimaging datasets revealed that faster DunedinPACNI predicted participants' cognitive impairment, accelerated brain atrophy, and conversion to diagnosed dementia. Underscoring close links between longitudinal aging of the body and brain, faster DunedinPACNI also predicted physical frailty, poor health, future chronic diseases, and mortality in older adults. Furthermore, DunedinPACNI followed the expected socioeconomic health gradient. When compared to brain age gap, an existing MRI aging biomarker, DunedinPACNI was similarly or more strongly related to clinical outcomes. DunedinPACNI is a "next generation" MRI measure that will be made publicly available to the research community to help accelerate aging research and evaluate the effectiveness of dementia prevention and anti-aging strategies.

BundleAGE: Predicting White Matter Age using Along-Tract Microstructural Profiles from Diffusion MRI

Brain Age Gap Estimation (BrainAGE) is an estimate of the gap between a person's chronological age (CA) and a measure of their brain's 'biological age' (BA). This metric is often used as a marker of accelerated aging, albeit with some caveats. Age prediction models trained on brain structural and functional MRI have been employed to derive BrainAGE biomarkers, for predicting the risk of neurodegeneration. While voxel-based and along-tract microstructural maps from diffusion MRI have been used to study brain aging, no studies have evaluated along-tract microstructure for computing BrainAGE. In this study, we train machine learning models to predict a person's age using along-tract microstructural profiles from diffusion tensor imaging. We were able to demonstrate differential aging patterns across different white matter bundles and microstructural measures. The novel Bundle Age Gap Estimation (BundleAGE) biomarker shows potential in quantifying risk factors for neurodegenerative diseases and aging, while incorporating finer scale information throughout white matter bundles.

Neuroanatomical and clinical factors predicting future cognitive impairment

Identifying cognitively normal (CN) older adults who will convert to cognitive impairment (CI) due to Alzheimer's disease is crucial for early intervention. Clinical and neuroimaging measures were acquired from 301 CN adults who converted to CI within 15 years of baseline, and 294 who did not. Regional volumes and brain age measures were extracted from T1-weighted magnetic resonance images. Linear discriminant analysis compared non-converters' characteristics against those of short-, mid-, and long-term converters. Conversion was associated with clinical measures such as hearing impairment and self-reported memory decline. Converters' brain volumes were smaller than non-converters' across 48 frontal, temporal, and subcortical structures. Brain age measures of 12 structures were correlated with shorter times to conversion. Conversion prediction accuracy increased from 81.5% to 90.5% as time to conversion decreased. Proximity to CI conversion is foreshadowed by anatomic features of brain aging that enhance the accuracy of predicting conversion.

Accelerated Aging after Traumatic Brain Injury: An ENIGMA Multi-Cohort Mega-Analysis

OBJECTIVE

The long-term consequences of traumatic brain injury (TBI) on brain structure remain uncertain. Given evidence that a single significant brain injury event increases the risk of dementia, brain-age estimation could provide a novel and efficient indexing of the long-term consequences of TBI. Brain-age procedures use predictive modeling to calculate brain-age scores for an individual using structural magnetic resonance imaging (MRI) data. Complicated mild, moderate, and severe TBI (cmsTBI) is associated with a higher predicted age difference (PAD), but the progression of PAD over time remains unclear. We sought to examine whether PAD increases as a function of time since injury (TSI) and if injury severity and sex interacted to influence this progression.

METHODS

Through the ENIGMA Adult Moderate and Severe (AMS)-TBI working group, we examine the largest TBI sample to date (n = 343), along with controls, for a total sample size of n = 540, to replicate and extend prior findings in the study of TBI brain age. Cross-sectional T1w-MRI data were aggregated across 7 cohorts, and brain age was established using a similar brain age algorithm to prior work in TBI.

RESULTS

Findings show that PAD widens with longer TSI, and there was evidence for differences between sexes in PAD, with men showing more advanced brain age. We did not find strong evidence supporting a link between PAD and cognitive performance.

INTERPRETATION

This work provides evidence that changes in brain structure after cmsTBI are dynamic, with an initial period of change, followed by relative stability in brain morphometry, eventually leading to further changes in the decades after a single cmsTBI. ANN NEUROL 2024;96:365-377.

An Attention-Based Hemispheric Relation Inference Network for Perinatal Brain Age Prediction

Brain anatomical age is an effective feature to assess the status of the brain, such as atypical development and aging. Although some deep learning models have been developed for estimating infant brain age, the performance of these models was unsatisfactory because few of them considered the developmental characteristics of brain anatomy during the perinatal period-the most rapid and complex developmental stage across the lifespan. The present study proposed an attention-based hemispheric relation inference network (HRINet) that takes advantage of the nature of brain structural lateralization during early development. This model captures the inter-hemispheric relationship using a graph attention mechanism and transmits lateralization information as features to describe the interactive development between bilateral hemispheres. The HRINet was used to estimate the brain age of 531 preterm and full-term neonates from the Developing Human Connectome Project (dHCP) database based on two metrics (mean curvature and sulcal depth) characterizing the folding morphology of the cortex. Our results showed that the HRINet outperformed other benchmark models in fitting the perinatal brain age, with mean absolute error of 0.53 and determination coefficient of 0.89. We also verified the generalizability of the HRINet on an extra independent dataset collected from the Gansu Provincial Maternity and Child-care Hospital. Furthermore, by applying the best-performing model to an independent dataset consisting of 47 scans of preterm infants at term-equivalent age, we showed that the predicted age was significantly lower than the chronological age, suggesting a delayed development of premature brains. Our results demonstrate the effectiveness and generalizability of the HRINet in estimating infant brain age, providing promising clinical applications for assessing neonatal brain maturity.

Brain Age Analysis and Dementia Classification using Convolutional Neural Networks trained on Diffusion MRI: Tests in Indian and North American Cohorts

Deep learning models based on convolutional neural networks (CNNs) have been used to classify Alzheimer's disease or infer dementia severity from T1-weighted brain MRI scans. Here, we examine the value of adding diffusion-weighted MRI (dMRI) as an input to these models. Much research in this area focuses on specific datasets such as the Alzheimer's Disease Neuroimaging Initiative (ADNI), which assesses people of North American, largely European ancestry, so we examine how models trained on ADNI, generalize to a new population dataset from India (the NIMHANS cohort). We first benchmark our models by predicting 'brain age' - the task of predicting a person's chronological age from their MRI scan and proceed to AD classification. We also evaluate the benefit of using a 3D CycleGAN approach to harmonize the imaging datasets before training the CNN models. Our experiments show that classification performance improves after harmonization in most cases, as well as better performance for dMRI as input.

Accelerated Brain Aging Mediates the Association Between Psychological Profiles and Clinical Pain in Knee Osteoarthritis

Chronic pain is driven by factors across the biopsychosocial spectrum. Previously, we demonstrated that magnetic resonance images (MRI)-based brain-predicted age differences (brain-PAD: brain-predicted age minus chronological age) were significantly associated with pain severity in individuals with chronic knee pain. We also previously identified four distinct, replicable, multidimensional psychological profiles significantly associated with clinical pain. The brain aging-psychological characteristics interface in persons with chronic pain promises elucidating factors contributing to their poor health outcomes, yet this relationship is barely understood. That is why we examined the interplay between the psychological profiles in participants having chronic knee pain impacting function, brain-PAD, and clinical pain severity. Controlling for demographics and MRI scanner, we compared the brain-PAD among psychological profiles at baseline (n = 164) and over two years (n = 90). We also explored whether profile-related differences in pain severity were mediated by brain-PAD. Brain-PAD differed significantly between profiles (ANOVA's omnibus test, P = .039). Specifically, participants in the profile 3 group (high negative/low positive emotions) had an average brain-PAD ∼4 years higher than those in profile- (low somatic reactivity), with P = .047, Bonferroni-corrected, and than those in profile 2 (high coping), with P = .027, uncorrected. Repeated measures ANOVA revealed no significant change in profile-related brain-PAD differences over time, but there was a significant decrease in brain-PAD for profile 4 (high optimism/high positive affect), with P = .045. Moreover, profile-related differences in pain severity at baseline were partly explained by brain-PAD differences between profile 3 and 1, or 2; but brain-PAD did not significantly mediate the influence of variations in profiles on changes in pain severity over time. PERSPECTIVE: Accelerated brain aging could underlie the psychological-pain relationship, and psychological characteristics may predispose individuals with chronic knee pain to worse health outcomes via neuropsychological processes.

Brain age prediction and deviations from normative trajectories in the neonatal connectome

Structural and functional connectomes undergo rapid changes during the third trimester and the first month of postnatal life. Despite progress, our understanding of the developmental trajectories of the connectome in the perinatal period remains incomplete. Brain age prediction uses machine learning to estimate the brain's maturity relative to normative data. The difference between the individual's predicted and chronological age-or brain age gap (BAG)-represents the deviation from these normative trajectories. Here, we assess brain age prediction and BAGs using structural and functional connectomes for infants in the first month of life. We used resting-state fMRI and DTI data from 611 infants (174 preterm; 437 term) from the Developing Human Connectome Project (dHCP) and connectome-based predictive modeling to predict postmenstrual age (PMA). Structural and functional connectomes accurately predicted PMA for term and preterm infants. Predicted ages from each modality were correlated. At the network level, nearly all canonical brain networks-even putatively later developing ones-generated accurate PMA prediction. Additionally, BAGs were associated with perinatal exposures and toddler behavioral outcomes. Overall, our results underscore the importance of normative modeling and deviations from these models during the perinatal period.

A Framework for Interpretability in Machine Learning for Medical Imaging

Interpretability for machine learning models in medical imaging (MLMI) is an important direction of research. However, there is a general sense of murkiness in what interpretability means. Why does the need for interpretability in MLMI arise? What goals does one actually seek to address when interpretability is needed? To answer these questions, we identify a need to formalize the goals and elements of interpretability in MLMI. By reasoning about real-world tasks and goals common in both medical image analysis and its intersection with machine learning, we identify five core elements of interpretability: localization, visual recognizability, physical attribution, model transparency, and actionability. From this, we arrive at a framework for interpretability in MLMI, which serves as a step-by-step guide to approaching interpretability in this context. Overall, this paper formalizes interpretability needs in the context of medical imaging, and our applied perspective clarifies concrete MLMI-specific goals and considerations in order to guide method design and improve real-world usage. Our goal is to provide practical and didactic information for model designers and practitioners, inspire developers of models in the medical imaging field to reason more deeply about what interpretability is achieving, and suggest future directions of interpretability research.

BrainAGE, brain health, and mental disorders: A systematic review

The imaging-based method of brainAGE aims to characterize an individual's vulnerability to age-related brain changes. The present study systematically reviewed brainAGE findings in neuropsychiatric conditions and discussed the potential of brainAGE as a marker for biological age. A systematic PubMed search (from inception to March 6th, 2023) identified 273 articles. The 30 included studies compared brainAGE between neuropsychiatric and healthy groups (n≥50). We presented results qualitatively and adapted a bias risk assessment questionnaire. The imaging modalities, design, and input features varied considerably between studies. While the studies found higher brainAGE in neuropsychiatric conditions (11 mild cognitive impairment/ dementia, 11 schizophrenia spectrum/ other psychotic and bipolar disorder, six depression/ anxiety, two multiple groups), the associations with clinical characteristics were mixed. While brainAGE is sensitive to group differences, limitations include the lack of diverse training samples, multi-modal studies, and external validation. Only a few studies obtained longitudinal data, and all have used algorithms built solely to predict chronological age. These limitations impede the validity of brainAGE as a biological age marker.

Application of AI in biological age prediction

The development of anti-aging interventions requires quantitative measurement of biological age. Machine learning models, known as "aging clocks," are built by leveraging diverse aging biomarkers that vary across lifespan to predict biological age. In addition to traditional aging clocks harnessing epigenetic signatures derived from bulk samples, emerging technologies allow the biological age estimating at single-cell level to dissect cellular diversity in aging tissues. Moreover, imaging-based aging clocks are increasingly employed with the advantage of non-invasive measurement, making it suitable for large-scale human cohort studies. To fully capture the features in the ever-growing multi-modal and high-dimensional aging-related data and uncover disease associations, deep-learning based approaches, which are effective to learn complex and non-linear relationships without relying on pre-defined features, are increasingly applied. The use of big data and AI-based aging clocks has achieved high accuracy, interpretability and generalizability, guiding clinical applications to delay age-related diseases and extend healthy lifespans.

eXplainable Artificial Intelligence (XAI) in aging clock models

XAI is a rapidly progressing field of machine learning, aiming to unravel the predictions of complex models. XAI is especially required in sensitive applications, e.g. in health care, when diagnosis, recommendations and treatment choices might rely on the decisions made by artificial intelligence systems. AI approaches have become widely used in aging research as well, in particular, in developing biological clock models and identifying biomarkers of aging and age-related diseases. However, the potential of XAI here awaits to be fully appreciated. We discuss the application of XAI for developing the "aging clocks" and present a comprehensive analysis of the literature categorized by the focus on particular physiological systems.

Brain-predicted age difference estimated using DeepBrainNet is significantly associated with pain and function-a multi-institutional and multiscanner study

ABSTRACT

Brain age predicted differences (brain-PAD: predicted brain age minus chronological age) have been reported to be significantly larger for individuals with chronic pain compared with those without. However, a debate remains after one article showed no significant differences. Using Gaussian Process Regression, an article provides evidence that these negative results might owe to the use of mixed samples by reporting a differential effect of chronic pain on brain-PAD across pain types. However, some remaining methodological issues regarding training sample size and sex-specific effects should be tackled before settling this controversy. Here, we explored differences in brain-PAD between musculoskeletal pain types and controls using a novel convolutional neural network for predicting brain-PADs, ie, DeepBrainNet. Based on a very large, multi-institutional, and heterogeneous training sample and requiring less magnetic resonance imaging preprocessing than other methods for brain age prediction, DeepBrainNet offers robust and reproducible brain-PADs, possibly highly sensitive to neuropathology. Controlling for scanner-related variability, we used a large sample (n = 660) with different scanners, ages (19-83 years), and musculoskeletal pain types (chronic low back [CBP] and osteoarthritis [OA] pain). Irrespective of sex, brain-PAD of OA pain participants was ∼3 to 4.7 years higher than that of CBP and controls, whereas brain-PAD did not significantly differ among controls and CBP. Moreover, brain-PAD was significantly related to multiple variables underlying the multidimensional pain experience. This comprehensive work adds evidence of pain type-specific effects of chronic pain on brain age. This could help in the clarification of the debate around possible relationships between brain aging mechanisms and pain.

Feasibility of brain age predictions from clinical T1-weighted MRIs

An individual's brain predicted age minus chronological age (brain-PAD) obtained from MRIs could become a biomarker of disease in research studies. However, brain age reports from clinical MRIs are scant despite the rich clinical information hospitals provide. Since clinical MRI protocols are meant for specific clinical purposes, performance of brain age predictions on clinical data need to be tested. We explored the feasibility of using DeepBrainNet, a deep network previously trained on research-oriented MRIs, to predict the brain ages of 840 patients who visited 15 facilities of a health system in Florida. Anticipating a strong prediction bias in our clinical sample, we characterized it to propose a covariate model in group-level regressions of brain-PAD (recommended to avoid Type I, II errors), and tested its generalizability, a requirement for meaningful brain age predictions in new single clinical cases. The best bias-related covariate model was scanner-independent and linear in age, while the best method to estimate bias-free brain ages was the inverse of a scanner-independent and quadratic in brain age function. We demonstrated the feasibility to detect sex-related differences in brain-PAD using group-level regression accounting for the selected covariate model. These differences were preserved after bias correction. The Mean-Average Error (MAE) of the predictions in independent data was ∼8 years, 2-3 years greater than reports for research-oriented MRIs using DeepBrainNet, whereas an R2 (assuming no bias) was 0.33 and 0.76 for the uncorrected and corrected brain ages, respectively. DeepBrainNet on clinical populations seems feasible, but more accurate algorithms or transfer-learning retraining is needed.

Toward MR protocol-agnostic, unbiased brain age predicted from clinical-grade MRIs

The difference between the estimated brain age and the chronological age ('brain-PAD') could become a clinical biomarker. However, most brain age models were developed for research-grade high-resolution T1-weighted MRIs, limiting their applicability to clinical-grade MRIs from various protocols. We adopted a dual-transfer learning strategy to develop a model agnostic to modality, resolution, or slice orientation. We retrained a convolutional neural network (CNN) using 6281 clinical MRIs from 1559 patients, among 7 modalities and 8 scanner models. The CNN was trained to estimate brain age from synthetic research-grade magnetization-prepared rapid gradient-echo MRIs (MPRAGEs) generated by a 'super-resolution' method. The model failed with T2-weighted Gradient-Echo MRIs. The mean absolute error (MAE) was 5.86-8.59 years across the other modalities, still higher than for research-grade MRIs, but comparable between actual and synthetic MPRAGEs for some modalities. We modeled the "regression bias" in brain age, for its correction is crucial for providing unbiased summary statistics of brain age or for personalized brain age-based biomarkers. The bias model was generalizable as its correction eliminated any correlation between brain-PAD and chronological age in new samples. Brain-PAD was reliable across modalities. We demonstrate the feasibility of brain age predictions from arbitrary clinical-grade MRIs, thereby contributing to personalized medicine.

A deep neural network estimation of brain age is sensitive to cognitive impairment and decline

The greatest known risk factor for Alzheimer's disease (AD) is age. While both normal aging and AD pathology involve structural changes in the brain, their trajectories of atrophy are not the same. Recent developments in artificial intelligence have encouraged studies to leverage neuroimaging-derived measures and deep learning approaches to predict brain age, which has shown promise as a sensitive biomarker in diagnosing and monitoring AD. However, prior efforts primarily involved structural magnetic resonance imaging and conventional diffusion MRI (dMRI) metrics without accounting for partial volume effects. To address this issue, we post-processed our dMRI scans with an advanced free-water (FW) correction technique to compute distinct FW-corrected fractional anisotropy (FAFWcorr) and FW maps that allow for the separation of tissue from fluid in a scan. We built 3 densely connected neural networks from FW-corrected dMRI, T1-weighted MRI, and combined FW+T1 features, respectively, to predict brain age. We then investigated the relationship of actual age and predicted brain ages with cognition. We found that all models accurately predicted actual age in cognitively unimpaired (CU) controls (FW: r=0.66, p=1.62×10-32; T1: r=0.61, p=1.45×10-26, FW+T1: r=0.77, p=6.48×10-50) and distinguished between CU and mild cognitive impairment participants (FW: p=0.006; T1: p=0.048; FW+T1: p=0.003), with FW+T1-derived age showing best performance. Additionally, all predicted brain age models were significantly associated with cross-sectional cognition (memory, FW: β=-1.094, p=6.32×10-7; T1: β=-1.331, p=6.52×10-7; FW+T1: β=-1.476, p=2.53×10-10; executive function, FW: β=-1.276, p=1.46×10-9; T1: β=-1.337, p=2.52×10-7; FW+T1: β=-1.850, p=3.85×10-17) and longitudinal cognition (memory, FW: β=-0.091, p=4.62×10-11; T1: β=-0.097, p=1.40×10-8; FW+T1: β=-0.101, p=1.35×10-11; executive function, FW: β=-0.125, p=1.20×10-10; T1: β=-0.163, p=4.25×10-12; FW+T1: β=-0.158, p=1.65×10-14). Our findings provide evidence that both T1-weighted MRI and dMRI measures improve brain age prediction and support predicted brain age as a sensitive biomarker of cognition and cognitive decline.

Brain Anatomy-Guided MRI Analysis for Assessing Clinical Progression of Cognitive Impairment with Structural MRI

Brain structural MRI has been widely used for assessing future progression of cognitive impairment (CI) based on learning-based methods. Previous studies generally suffer from the limited number of labeled training data, while there exists a huge amount of MRIs in large-scale public databases. Even without task-specific label information, brain anatomical structures provided by these MRIs can be used to boost learning performance intuitively. Unfortunately, existing research seldom takes advantage of such brain anatomy prior. To this end, this paper proposes a brain anatomy-guided representation (BAR) learning framework for assessing the clinical progression of cognitive impairment with T1-weighted MRIs. The BAR consists of a pretext model and a downstream model, with a shared brain anatomy-guided encoder for MRI feature extraction. The pretext model also contains a decoder for brain tissue segmentation, while the downstream model relies on a predictor for classification. We first train the pretext model through a brain tissue segmentation task on 9,544 auxiliary T1-weighted MRIs, yielding a generalizable encoder. The downstream model with the learned encoder is further fine-tuned on target MRIs for prediction tasks. We validate the proposed BAR on two CI-related studies with a total of 391 subjects with T1-weighted MRIs. Experimental results suggest that the BAR outperforms several state-of-the-art (SOTA) methods. The source code and pre-trained models are available at https://github.com/goodaycoder/BAR.

Repurposing electroencephalogram monitoring of general anaesthesia for building biomarkers of brain ageing: an exploratory study

Background

Electroencephalography (EEG) is increasingly used for monitoring the depth of general anaesthesia, but EEG data from general anaesthesia monitoring are rarely reused for research. Here, we explored repurposing EEG monitoring from general anaesthesia for brain-age modelling using machine learning. We hypothesised that brain age estimated from EEG during general anaesthesia is associated with perioperative risk.

Methods

We reanalysed four-electrode EEGs of 323 patients under stable propofol or sevoflurane anaesthesia to study four EEG signatures (95% of EEG power <8-13 Hz) for age prediction: total power, alpha-band power (8-13 Hz), power spectrum, and spatial patterns in frequency bands. We constructed age-prediction models from EEGs of a healthy reference group (ASA 1 or 2) during propofol anaesthesia. Although all signatures were informative, state-of-the-art age-prediction performance was unlocked by parsing spatial patterns across electrodes along the entire power spectrum (mean absolute error=8.2 yr; R2=0.65).

Results

Clinical exploration in ASA 1 or 2 patients revealed that brain age was positively correlated with intraoperative burst suppression, a risk factor for general anaesthesia complications. Surprisingly, brain age was negatively correlated with burst suppression in patients with higher ASA scores, suggesting hidden confounders. Secondary analyses revealed that age-related EEG signatures were specific to propofol anaesthesia, reflected by limited model generalisation to anaesthesia maintained with sevoflurane.

Conclusions

Although EEG from general anaesthesia may enable state-of-the-art age prediction, differences between anaesthetic drugs can impact the effectiveness and validity of brain-age models. To unleash the dormant potential of EEG monitoring for clinical research, larger datasets from heterogeneous populations with precisely documented drug dosage will be essential.

Significant Acceleration of Regional Brain Aging and Atrophy After Mild Traumatic Brain Injury

Brain regions' rates of age-related volumetric change after traumatic brain injury (TBI) are unknown. Here, we quantify these rates cross-sectionally in 113 persons with recent mild TBI (mTBI), whom we compare against 3 418 healthy controls (HCs). Regional gray matter (GM) volumes were extracted from magnetic resonance images. Linear regression yielded regional brain ages and the annualized average rates of regional GM volume loss. These results were compared across groups after accounting for sex and intracranial volume. In HCs, the steepest rates of volume loss were recorded in the nucleus accumbens, amygdala, and lateral orbital sulcus. In mTBI, approximately 80% of GM structures had significantly steeper rates of annual volume loss than in HCs. The largest group differences involved the short gyri of the insula and both the long gyrus and central sulcus of the insula. No significant sex differences were found in the mTBI group, regional brain ages being the oldest in prefrontal and temporal structures. Thus, mTBI involves significantly steeper regional GM loss rates than in HCs, reflecting older-than-expected regional brain ages.

The capacity of action observation to drag the trainees' motor pattern toward the observed model

Action Observation Training (AOT) promotes the acquisition of motor abilities. However, while the cortical modulations associated with the AOT efficacy are well known, few studies investigated the AOT peripheral neural correlates and whether their dynamics move towards the observed model during the training. We administered seventy-two participants (randomized into AOT and Control groups) with training for learning to grasp marbles with chopsticks. Execution practice was preceded by an observation session, in which AOT participants observed an expert performing the task, whereas controls observed landscape videos. Behavioral indices were measured, and three hand muscles' electromyographic (EMG) activity was recorded and compared with the expert. Behaviorally, both groups improved during the training, with AOT outperforming controls. The EMG trainee-model similarity also increased during the training, but only for the AOT group. When combining behavioral and EMG similarity findings, no global relationship emerged; however, behavioral improvements were "locally" predicted by the similarity gain in muscles and action phases more related to the specific motor act. These findings reveal that AOT plays a magnetic role in motor learning, attracting the trainee's motor pattern toward the observed model and paving the way for developing online monitoring tools and neurofeedback protocols.

Structural-functional connectivity bandwidth predicts processing speed in mild traumatic brain Injury: A multiplex network analysis

An emerging body of work has revealed alterations in structural (SC) and functional (FC) brain connectivity following mild TBI (mTBI), with mixed findings. However, these studies seldom integrate complimentary neuroimaging modalities within a unified framework. Multilayer network analysis is an emerging technique to uncover how white matter organization enables functional communication. Using our novel graph metric (SC-FC Bandwidth), we quantified the information capacity of synchronous brain regions in 53 mild TBI patients (46 females; age mean = 40.2 years (y), σ = 16.7 (y), range: 18-79 (y). Diffusion MRI and resting state fMRI were administered at the acute and chronic post-injury intervals. Moreover, participants completed a cognitive task to measure processing speed (30 Seconds and Counting Task; 30-SACT). Processing speed was significantly increased at the chronic, relative to the acute post-injury intervals (p = <0.001). Nonlinear principal components of direct (t = -1.84, p = 0.06) and indirect SC-FC Bandwidth (t = 3.86, p = <0.001) predicted processing speed with a moderate effect size (R2 = 0.43, p < 0.001), while controlling for age. A subnetwork of interhemispheric edges with increased SC-FC Bandwidth was identified at the chronic, relative to the acute mTBI post-injury interval (pFDR = 0.05). Increased interhemispheric SC-FC Bandwidth of this network corresponded with improved processing speed at the chronic post-injury interval (partial r = 0.32, p = 0.02). Our findings revealed that mild TBI results in complex reorganization of brain connectivity optimized for maximum information flow, supporting improved cognitive performance as a compensatory mechanism. Moving forward, this measurement may complement clinical assessment as an objective marker of mTBI recovery.

Biomarkers of aging

Aging biomarkers are a combination of biological parameters to (i) assess age-related changes, (ii) track the physiological aging process, and (iii) predict the transition into a pathological status. Although a broad spectrum of aging biomarkers has been developed, their potential uses and limitations remain poorly characterized. An immediate goal of biomarkers is to help us answer the following three fundamental questions in aging research: How old are we? Why do we get old? And how can we age slower? This review aims to address this need. Here, we summarize our current knowledge of biomarkers developed for cellular, organ, and organismal levels of aging, comprising six pillars: physiological characteristics, medical imaging, histological features, cellular alterations, molecular changes, and secretory factors. To fulfill all these requisites, we propose that aging biomarkers should qualify for being specific, systemic, and clinically relevant.

Multifactorial Modeling of Cognitive Trajectories Using an Advanced Regression Technique: Improving Our Understanding of Biomarkers and Modifiable Variables that Support Cognition

Current research trends emphasize complex models of cognitive outcomes, with multiple, interacting predictors, including factors amenable to interventions toward sustaining healthy cognitive aging. Such models often require advanced analysis techniques. The article by Stark et al., 'Partial least squares regression analysis of Alzheimer's disease biomarkers, modifiable health variables, and cognitive change in older adults with mild cognitive impairment', uses partial least squares regression to examine the associations to memory and executive change of 29 biomarker and demographic variables. This commentary discusses the significance of their results and methods within the context of current research foci.