Neuroimaging-Based Brain Age Estimation: A Promising Personalized Biomarker in Neuropsychiatry

It is now possible to estimate an individual’s brain age via brain scans and machine-learning models. This validated technique has opened up new avenues for addressing clinical questions in neurology, and, in this review, we summarize the many clinical applications of brain-age estimation in neuropsychiatry and general populations. We first provide an introduction to typical neuroimaging modalities, feature extraction methods, and machine-learning models that have been used to develop a brain-age estimation framework. We then focus on the significant findings of the brain-age estimation technique in the field of neuropsychiatry as well as the usefulness of the technique for addressing clinical questions in neuropsychiatry. These applications may contribute to more timely and targeted neuropsychiatric therapies. Last, we discuss the practical problems and challenges described in the literature and suggest some future research directions.


Aging, Disease, and the Brain
The aging process in humans is associated with the progressive decline of various physiological and organ functions [1], and many diseases including cancer, cardiovascular disease, diabetes, and dementia are associated with aging [2]. It is not uncommon for elderly people to suffer from multiple diseases simultaneously. Since humans' bodies change with age and as humans are living longer in several regions of the world, the aging process has become a key issue in public health, disease prevention, and treatment. Many discussions concerning the pathological meaning of aging in the context of epigenetic change, proteotoxic or oxidative stress, and telomere damage have thus been conducted [3].
The brain is also affected by aging [4]. In the early stage of life, the aging process is regarded as brain development in which the brain matures, and children usually experience an increase in their cognitive ability along with their physical growth. During late adulthood, the brain aging process has different effects, e.g., a decline of cognitive function, and advancing age is associated with neurodegeneration, particularly Alzheimer's disease and other forms of dementia [5,6]. If the aging process of the brain could be measured precisely and accurately, the findings may have potential as biomarkers for neuropsychiatric disorders. In fact, frameworks to quantify the age of a human brain have been attempted for several decades [7]. Today, advances in medical imaging and analytical methods (especially machine learning) have allowed the calculation of an individual's biological age from the extracted biological features [8]. The frameworks that are now used to estimate the age of an individual's brain have the potential to provide useful, objective, and personalized biomarkers for neurological and psychiatric disorders.

Neuroimaging-Based Brain-Age Estimation
Telomere-related and epigenetics-related biomarkers have not shown sufficient predictive and deterministic value for estimating brain ages, and it has been suggested that phenotype-based estimation can generate a much closer indicator of brain age [4]. Neuroimaging is a widely available, less-invasive method to investigate the whole brain of humans, and with neuroimaging, the brain's morphological and microstructural features can be obtained; these features are speculated to be suitable material for the estimation of the age of an individual brain. In fact, neuroimaging-based brain-age estimation has been increasingly applied to individuals with various neuropsychiatric disorders and general populations [8]. In this narrative review, we examined over 100 studies and introduce the recent findings and methodologies of this emerging technique. We conducted a search of the PubMed database in May 2022 using "brain-age estimation" and/or "brain-age prediction" as keywords, although we did not adopt rigorous systematic selection criteria of studies for this narrative review.

Theory of Neuroimaging-Based Brain-Age Estimation
In 2010, Katja Franke and her peers developed a prediction model that was able to estimate a subject's age based on brain imaging data and the use of a regression machinelearning model [9]. The output of a brain-age estimation framework has been called the "brain age-delta," "brain predicted age difference (Brain-PAD)," "brain age gap estimation (BrainAGE)," and "brain age gap (BAG)," each of which is computed by deducting the estimated brain age from the subject's chronological age. In this review, we refer to "brain age-delta" as the output of a brain-age estimation framework. The brain age-delta is known as a heritable biomarker for both monitoring cognitively healthy aging and identifying age-associated disorders [8]. There are three possibilities for a brain age-delta value: (i) a brain age-delta close to zero, representing normal brain aging, (ii) a positive brain age-delta (i.e., estimated brain age > chronological age), representing an older-appearing brain, and (iii) a negative brain age-delta (i.e., estimated brain age < chronological age), representing a younger-appearing brain.
A brain-age estimation study is generally composed of three main stages: (i) creating a prediction model by using extracted brain features and a regression machine-learning model, validation, and bias correction; (ii) computing the brain age and brain age-delta for the subject under study; and (iii) interpreting the results, including the use of a within-group and/or a between-groups analysis. Figure 1 depicts the pipeline of a typical brain-age estimation study.

Input Data and Feature-Extraction Methodologies of Neuroimaging
One of the key concerns among researchers attempting to develop a brain-age estimation framework is the selection of the input data. Each modality offers unique insights into the brain. For example, fluorodeoxyglucose-positron emission tomography (FDG- In the literature, the typical accuracy of the prediction of brain ages is from 2 years to 10 years in terms of mean absolute error (MAE) [10,11]. The prediction accuracy in a brain-age estimation framework depends on variables such as the type of input data, the feature extraction, and reduction strategies [12] and bias adjustment techniques [13], and machine-learning models [14]. In the following subsections, we provide a general overview of these variables.

Data Reduction, Validation, and Bias Adjustment Neuroimaging Methodologies
The 'curse of dimensionality' is one of the major concerns in developing a brain-age estimation framework, particularly when the number of brain features is far higher than the number of samples (e.g., in voxel-based feature extraction strategies). High-dimensional data can give rise to some substantial issues in a prediction model, such as overfitting and decreased computational efficiency. A data reduction technique that can decrease the high dimensionality of data and diminish redundant information is thus required. In the area of brain-age estimation, most studies have used the principal component analysis (PCA) strategy, which is an unsupervised learning technique [9,19]. The effect of the number of principal components on the accuracy of brain-age predictions has been investigated [9].
The number of principal components may influence the prediction accuracy in a brain age estimation framework. However, it can be adjusted to achieve maximum accuracy in the training set [9]. After a prediction model is developed, it is critical to validate the model's prediction accuracy. Most studies in the field of brain-age estimation have used a K-fold crossvalidation strategy (e.g., K = 5 or 10) to assess the prediction performance on a training set [12,14,16,21]. In the K-fold cross-validation technique, the data are randomly divided into K folds, and the learning process is repeated K times so that K-1 folds are used for training a prediction model, and the remaining fold is used as a test for each iteration. To assess the prediction accuracy, researchers generally use the coefficient of determination (R2) between the subjects' chronological age and estimated age, the MAE, and root mean square error (RMSE) metrics. Many brain-age estimation studies have reported age dependency on the prediction outputs, and this is considered a substantial issue in brain-age frameworks [13,21]. This bias, which could be a result of regression dilution bias, may adversely affect the predicted values and alter the interpretation of results. Several techniques have been proposed to diminish this bias (i.e., age dependency) [13,21,25]. For instance, Le and colleagues proposed using chronological age as a covariate in the statistical analyses and interpreting the results [26]. However, it should be highlighted that Le's method is appropriate for group comparison only and not able to deliver bias-free brain age values at the individual level. A bias adjustment strategy is proposed in [21] (i.e., Cole's method) that uses the intercept and slope of a linear regression model of estimated brain age against chronological derived from the training set. The bias-free Brain-age values in the test sets are then calculated by subtracting the intercept from the predicted brain age and dividing by the slope [21]. The most recent bias adjustment technique is suggested in [13] (i.e., Beheshti's method) which computes offset values for test subjects on the basis of the intercept and slope of a linear regression model of brain age-delta against chronological age achieved from the training set. Then, the bias-free Brain-age values are computed by subtracting the offset values from the estimated brain-age values [13]. A direct comparison of these bias adjustment techniques has shown that Beheshti's method greatly reduces the variance of the predicted ages, whereas Cole's method increases it [13].

Machine-Learning Methodologies
One of the important steps in developing a brain-age estimation framework is choosing a regression machine-learning model. A regression model establishes a pattern between independent variables (here, brain features) and the corresponding dependent variable (a subject's chronological age) based on the training dataset, and the model uses this pattern to predict the brain age based on unseen data (i.e., independent test datasets). The most widely used traditional regression algorithms include support vector regression (SVR) [19,23], relevance vector regression (RVR) [9], Gaussian process regression [21], an ensemble of gradient-boosted regression trees [25], and XGBoost [25]. It has been demonstrated that the type of regression algorithm used influences the prediction accuracy and the interpretation of outcomes in brain-age frameworks [14].
In addition to the traditional regression algorithms, deep learning models have become a prominent methodology in the area of brain-age estimation [11], as they can be used to develop more accurate prediction models. A major advantage of deep learning models is that they can be applied directly with 3D brain image data and incorporate feature extraction, data reduction, and prediction stages into a unified system. The main challenge of deep learning-based brain-age estimation frameworks is that this methodology requires a large dataset to train a model. In 2017, James Cole and his peers developed the first deep learning-based brain-age estimation framework, with a 3D convolutional neural network (CNN) that uses 3D gray matter and 3D white matter intensity maps as the input data [11]. Other deep learning architectures used in brain-age estimation frameworks include feed-forward neural networks, VGGNet [27], ResNet [28], U-Net [29], and an ensemble of CNN architectures [30].

Alzheimer's Disease, Dementia, and Memory Impairment
One of the most active areas of brain-age research concerns Alzheimer's disease (AD) and mild cognitive impairment (MCI) ( Table 1) [17,23,[31][32][33][34][35][36][37], because of their strong association with aging. Alzheimer's disease is the most common cause of dementia, which is also a relevant issue in aging societies in many developed countries. The early detection of AD is important in terms of early cognitive intervention [38][39][40] as well as the future development of disease-modifying therapy [41], and neuroimaging plays key roles in ensuring accurate and early diagnoses, revealing the underlying pathophysiology, and monitoring the disease. An increased BAG in individuals with AD has been consistently reported, ranging from +2.88 to +9.29 years [17,20,37], and correlations between an increased BAG and cognitive dysfunction or white matter hyperintensity were also found [17,36]. Importantly, the predictability of progression from MCI to AD and the detectability of preclinical AD based on brain-age measures are also confirmed and would be clinically significant [31,34,37].

Other Neurological Diseases
Parkinson's disease (PD) is a common neurodegenerative movement disorder characterized by the degeneration of dopaminergic neurons in the substantia nigra [55]. Overall, it seems that an increase in brain age by 2-3 years occurs in PD [20,42,43], and such an increase is associated with cognitive or motor impairment (Table 1). A comparison study between PD and AD revealed a significant increase in the BAG in AD compared to PD.
Epilepsy is also a common neurological disorder, characterized by recurrent seizures associated with abnormal electrical activity in the brain. A brain with chronic epilepsy tends to present a BAG of +4 to +8 years (Table 1) [19,[44][45][46], and comorbid psychosis may further increase the BAG by several additional years [19]. Interestingly, epilepsy surgery may reduce the abnormal BAG increase, regardless of postsurgical seizure freedom [46].
Multiple sclerosis (MS) is an autoimmune disease involving damage to the myelin sheaths of the brain and spinal cord [56]. The BAG in MS is relatively high at +6.5 to 10.3 years on average (Table 1), and it is particularly higher in secondary progressive MS (+13.3 years) [47,48]. An increased BAG is also suggested to predict MS progression. A brain-age framework has also been applied to neurological and related disorders including traumatic brain injury [49,50], pain [51,52], Prader-Willi syndrome [53], and HIV infection [54] (Table 1).

Schizophrenia and Psychotic Disorders
Schizophrenia is a serious psychiatric disorder presenting symptoms that include psychosis, cognitive dysfunction, and negative symptoms. Increased brain age in schizophrenia and first-episode psychosis (FEP) has been reported ( Table 2) [57][58][59][60][61][62][63][64][65][66][67][68][69][70][71][72], and consistent findings of a BAG increased by approx. 3-6 years in schizophrenia have been confirmed, with possible associations with cognitive dysfunction or polygenic risk. The BAG in individuals with FEP may be lower than that in schizophrenia, and, according to longitudinal studies, an acceleration of brain aging over time is suggested in this population. The BAG is also associated with schizotypal symptoms in relatives of patients with psychosis [72]. Early medication may reduce the BAG in psychosis [71].

Other Psychiatric Disorders
The brain age in other psychiatric disorders such as obsessive-compulsive disorder (OCD) and specific phobias has been investigated [82,83] (Table 2), and a relatively large study reported contributions of both severe mental illness and cardiometabolic disorders to an increased BAG [84].

Comprehensive Studies
Brain-age findings across various neuropsychiatric disorders have been obtained in comprehensive studies [85][86][87][88] (Table 2). Overall, these studies successfully identified neuropsychiatric disorders and risk factors by using brain age, and it was indicated that a multimodal imaging model may have high accuracy [88]. In particular, an investigation of a large sample (>10,000 patients and 35,000 healthy controls) revealed the effect sizes of a BAG in various conditions, which should be regarded as a reliable standard of BAG scores so far [85]. According to this study, the strongest aging of the brain is seen in dementia, followed by MS, schizophrenia, and MCI.

Applications to General Populations
Targeting a general population or individuals without neuropsychiatric diagnoses is another important topic in neuroimaging-based brain-age framework research, as it may clarify how to keep our brains healthy and avoid the risks of accelerated aging [15,16,21,. The most consistent significant risk factor could be diabetes (Table 3). In fact, diabetes has been consistently reported to adversely affect the aging of the brain. Alcohol consumption and smoking were also associated with an increased BAG in some studies [15,16,[121][122][123]128]. Other factors that were suggested to be associated with an increased BAG include mortality, allostatic load, lung function, exposure to famine in early gestation, recidivism, chronic pain, cardiovascular risk, chemotherapy for cancer, lead exposure in childhood, hypertension, premature birth, male sex, worry and rumination, neighborhood disadvantage, sleep apnea, obesity, and physical strength (Table 3). BAG: +4.3 yr in males whose mothers were exposed to famine in early gestation   In addition, several studies reported potentially protective factors associated with a reduced BAG: long-term meditation (−7.5 years), music composition (approx. −4 years), physical activity, taking ibuprofen, and life satisfaction (Table 3). Interestingly, it has been observed in more than one study that childbirth decreases the BAG in women, not only during the postpartum period but also in later life [100].
Thus, although the studies are diverse in terms of the methodologies used and the targeted factors, cumulative evidence will further expand our knowledge of how to improve the aging process of human brains. The strong and consistent risk for brain aging appears to be diabetes, followed by alcohol consumption. Other lifestyle-related risk factors, e.g., smoking or hypertension, may also be harmful but less consistent. Regarding beneficial effects, though the research focuses were diverse across studies, physical, mental, or creative activities may likely improve our brain age. It is unclear whether neuropsychiatric disorders, particularly dementia, could be prevented by improving brain aging. Further research is necessary to obtain real-world evidence regarding the utility of brain-age studies for this question.

Strengths, Controversies, and Future Direction
As described above, a neuroimaging-based brain-age estimation can provide a reliable neuropsychiatric biomarker at the single-subject level. In addition, brain MRI-particularly T1-weighted structural MRI-is a widely available examination in most countries, which may support easier and wider clinical applications of brain-age analyses. Given that many studies have successfully used public databases to build a brain-age prediction model, the reproducibility and external validity of a brain-age model should be at an acceptable level. Thus, the strengths of brain age as a biomarker would be its use as a single-subject-level marker, widely available examination, and acceptable reproducibility. The processing of MRI scans, including normalization and machine-learning analysis, may require advanced techniques and could be a possible barrier for most facilities, but currently, there are several public tools, e.g., BARACUS (https://github.com/BIDS-Apps/baracus, accessed on 31 October 2022) and brainageR (https://github.com/james-cole/brainageR, accessed on 31 October 2022) which would help us apply a brain age model to the patients.
Controversies and limitations of the use of brain ages have also been suggested, particularly for the interpretation of study results. As reviewed herein, there is a level of overlapping of BAG scores across various disorders, which might limit the usefulness of the brain age for differential diagnoses in clinical analyses. It was also reported that individual variations in brain age were associated with early-life factors rather than longitudinal changes [119]. Moreover, the methodology is quite diverse in terms of imaging modalities, processing, and choice of machine-learning algorithm, and there is no established consensus about the optimal protocol for determining brain ages. Future studies should address these controversies and limitations.
In conclusion, neuroimaging-based brain-age estimation has been widely and increasingly researched for over 10 years, and many studies revealed its usefulness for neuropsychiatry. Considering the utility, availability, and reproducibility of neuroimagingbased brain-age estimations for single patients, brain age can be expected to become a useful personalized biomarker in neuropsychiatry.
Author Contributions: Both D.S. and I.B. contributed equally to this article including searching for literature and writing a manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This study was supported by grants from the Japan Society for the Promotion of Science (KAKENHI) no. JP21K15720, the Japan Epilepsy Research Foundation (JERF TENKAN 22007), and the Uehara Memorial Foundation (all to D.S.).