Functional Brain Age Acceleration from Dynamic and Static Connectivity Predicts Working Memory and Attention Deficits in Schizophrenia

Functional Brain Age Acceleration from Dynamic and Static Connectivity Predicts Working Memory and Attention Deficits in Schizophrenia | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Functional Brain Age Acceleration from Dynamic and Static Connectivity Predicts Working Memory and Attention Deficits in Schizophrenia Sabrina J. Edwards-Swart , Bradley Baker , Daniel H. Mathalon , Judith M. Ford , Adrian Preda , Theo G.M. van Erp , Godfrey D. Pearlson , Jessica A. Turner , View ORCID Profile Vince D. Calhoun , Mohammad S. E. Sendi doi: https://doi.org/10.1101/2025.07.21.25331956 Sabrina J. Edwards-Swart 1 Department of Computer Science, Georgia Institute of Technology , Atlanta, GA, United States 2 Tri-institutional Center for Translational Research in Neuroimaging and Data Science: Georgia State University, Georgia Institute of Technology, Emory University Atlanta , GA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Bradley Baker 2 Tri-institutional Center for Translational Research in Neuroimaging and Data Science: Georgia State University, Georgia Institute of Technology, Emory University Atlanta , GA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Daniel H. Mathalon 3 Department of Psychiatry, Weill Institute of Neurosciences, University of California , San Francisco, CA, United States 4 Mental Health Service, Veterans Affairs San Francisco Healthcare System , San Francisco, CA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Judith M. Ford 3 Department of Psychiatry, Weill Institute of Neurosciences, University of California , San Francisco, CA, United States 4 Mental Health Service, Veterans Affairs San Francisco Healthcare System , San Francisco, CA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Adrian Preda 5 Department of Psychiatry and Human Behavior, University of California , Irvine, Irvine, CA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Theo G.M. van Erp 5 Department of Psychiatry and Human Behavior, University of California , Irvine, Irvine, CA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Godfrey D. Pearlson 6 Department of Psychiatry, School of Medicine, Yale University , New Haven, CT, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jessica A. Turner 7 Department of Psychiatry and Behavioral Health, College of Medicine, The Ohio State University , Columbus, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Vince D. Calhoun 2 Tri-institutional Center for Translational Research in Neuroimaging and Data Science: Georgia State University, Georgia Institute of Technology, Emory University Atlanta , GA, United States 8 Department of Computer Science, Georgia State University , Atlanta, GA, United States 9 Department of Electrical and Computer Engineering, Georgia Institute of Technology , Atlanta, GA, United States 10 Department of Psychology, Neuroscience Institute, Georgia State University , Atlanta, GA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Vince D. Calhoun Mohammad S. E. Sendi 2 Tri-institutional Center for Translational Research in Neuroimaging and Data Science: Georgia State University, Georgia Institute of Technology, Emory University Atlanta , GA, United States 11 Division of Depression and Anxiety, McLean Hospital , Belmont, MA, United States USA 12 Department of Psychiatry, Harvard Medical School , Boston, MA, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: msendi{at}mclean.harvard.edu vcalhoun{at}gsu.edu Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF ABSTRACT Background Schizophrenia is characterized by deficits in attention and working memory. In recent years, the brain age gap (BAG), defined as the difference between neuroimaging-predicted and chronological age, has emerged as a biomarker of brain dysfunction. Prior studies primarily use structural MRI or static functional network connectivity (sFNC), while the potential of dynamic functional network connectivity (dFNC) to quantify BAG in relationship with cognition remains underexplored. Methods Leveraging one of the largest resting-state fMRI datasets to date (N=22,569; UK Biobank, HCP, HCP-Aging), we developed robust brain age prediction models incorporating both wide-brain and sub-network variables derived from sFNC and dFNC. These models were validated in an independent clinical sample (FBIRN; N=153) including individuals with schizophrenia and healthy controls. Associations between BAGs and cognitive measures (attention vigilance, working memory) were evaluated using general linear models, controlling for key demographic and clinical covariates. Results Both sFNC and dFNC models demonstrated robust prediction accuracy in healthy individuals (sFNC: r =0.8755; dFNC: r =0.8675). Wide-brain BAGs (wBAGs) showed strong negative associations with attention vigilance (sFNC: r =–0.2923, FDR p =0.0013; dFNC: r =– 0.2715, FDR p =0.0016) and working memory (sFNC: r =–0.2237, FDR p =0.0088; dFNC: r =– 0.2508, FDR p =0.0064). Sub-network BAGs (subBAGs) within subcortical, sensorimotor, cognitive control, and default mode networks robustly predicted cognitive deficits, with dFNC-derived subBAGs showing the strongest effects ( FDR p <0.01). Conclusions Our findings establish dFNC-based BAGs as a sensitive and clinically relevant biomarker of cognitive impairment in schizophrenia, outperforming sFNC and highlighting the translational potential of dynamic connectivity for precision diagnosis and treatment. 1. INTRODUCTION Schizophrenia, affecting 0.25% to 0.64% of the U.S. population, is a complex psychiatric disorder marked by profound disruptions in cognition, emotion, and social interactions ( 1 , 2 ). Key symptoms include hallucinations, delusions, and motor impairments, alongside significant reductions in emotional expression and social engagement. Critically, individuals with schizophrenia exhibit marked cognitive deficits, particularly in working memory and attention vigilance. Studies indicate they perform approximately 2.5 standard deviations below healthy controls in these cognitive domains ( 3 ). These deficits are characterized by difficulties in retaining and updating information in working memory ( 4 ), with increasing cognitive demands leading to more pronounced declines in attention vigilance compared to healthy individuals ( 5 ). In recent decades, neuroimaging techniques have greatly enhanced our understanding of cognitive function and attention deficits in schizophrenia ( 4 , 6 ). Structural magnetic resonance imaging (sMRI) provides detailed anatomical insights ( 7 ), revealed associations between executive function impairments and reductions in prefrontal cortex volume and thickness, and linked episodic memory deficits with hippocampal atrophy ( 8 ). Conversely, functional magnetic resonance imaging (fMRI) measures brain activity through oxygen metabolism and blood flow, illuminated functional dynamics across various brain regions ( 9 ). Functional MRI has been pivotal in identifying brain regions involved in working memory, offering valuable insights into cognitive impairments in patient populations compared to healthy controls. Broadly, Neuroimaging plays a key role in examining executive function and episodic memory impairments and provides a valuable tool for tracking brain function changes following cognitive remediation therapies ( 8 ). In recent years, brain age prediction from neuroimaging has emerged as a promising approach for identifying novel biomarkers of neurological and neuropsychiatric conditions ( 14 , 24 , 26 , 47 , 49 , 51 ). By modeling the relationship between neuroimaging variables and age in healthy individuals, this method enables the detection of meaningful brain deviations in patient populations. The difference between an individual’s predicted brain age and their chronological age, termed the brain age gap (BAG), provides critical insights into the pathology of cognitive function ( 10 – 13 ). For example, emerging evidence demonstrates that individuals with accelerated brain age consistently show marked deficits across critical cognitive domains— including IQ, verbal comprehension, perceptual reasoning, processing speed, working memory, and memory recall—as captured by gold-standard assessments such as the Rey Auditory Verbal Learning Test ( 12 ). In schizophrenia, studies using sMRI have revealed that patients often exhibit a brain age 6 to 8 years older than their chronological age ( 15 ). Further, the BAG was noted to increase across at-risk, recent onset ( 16 ), and recurrent phases of schizophrenia, initially rising approximately a year and a half annually for the first five years before stabilizing ( 17 ). Additionally, early-stage schizophrenia and bipolar disorder present distinct BAG profiles. Individuals with schizophrenia consistently exhibit elevated BAGs, whereas those with bipolar disorder maintain levels comparable to healthy controls. This distinction underscores the potential of BAG as a valuable biomarker for enhancing early differential diagnosis ( 18 ). This suggests a progressive pathogenic component unique to schizophrenia, underscoring the utility of BAG as a biomarker ( 19 ). Moreover, a significant negative association between BAG and cognitive functions like working memory has been observed in schizophrenia, highlighting the potential of BAG in understanding and diagnosing declined neurocognitive performance in schizophrenia ( 20 ). While the majority of brain age prediction research has traditionally focused on structural MRI (sMRI), recent advancements include employing functional connectivity metrics from resting-state fMRI to predict brain age ( 21 – 24 ). A notable study using resting-state fMRI data from the Philadelphia Neurodevelopmental Cohort demonstrated the value of brain age models in youth by linking older estimated brain age to greater symptom burden, especially across DSM-5 psychiatric diagnoses. These findings underscore the potential of brain age to capture deviations in brain maturation associated with mental health conditions ( 25 ). Furthermore, an innovative study used resting-state fMRI and brain age prediction to identify neural connections associated with abnormal brain aging. By systematically excluding connections from the training model, the approach pinpointed those most critical to brain age accuracy, providing new insights into the neurobiological mechanisms of age-related psychiatric conditions ( 26 ). Although brain age prediction using static functional connectivity (sFNC), estimated from resting-state fMRI, has proven valuable, the potential of dynamic functional connectivity (dFNC) in this area is still underexplored. Unlike sFNC, which evaluates connectivity from correlations across an entire time series, dFNC analyzes connections between brain regions or networks within specific time intervals. This allows for the capture of temporal fluctuations that reveal how brain network interactions change over time ( 27 , 28 ). The variability and dynamism of neural signals, captured by dFNC, are crucial in understanding cognitive deficits and clinical symptoms in psychiatric disorders. These dynamic measures provide insights into brain function that static approaches cannot, highlighting changes and interactions that are critical in disease progression ( 29 – 35 ). Because cognitive symptoms in schizophrenia are potentially driven by disrupted brain network dynamics, leveraging a BAG derived from dFNC offers a powerful, next-generation marker of disease-related neurobiological aging. Critically, no study has yet evaluated whether dFNC-based BAG outperforms traditional static FNC (sFNC) BAG in explaining core cognitive deficits such as working memory and attention. Establishing this added predictive power would mark a transformative advance—positioning dFNC BAG as a mechanistically grounded biomarker and unlocking new precision targets for circuit-based interventions in schizophrenia and related psychiatric conditions. Here we address this gap using the largest sample to date: 22,569 resting-state scans from 17□522 healthy adults from UK Biobank ( 36 , 37 ) and Human Connectome Project ( 38 – 40 ) to train multiple wide-brain and sub-network brain age prediction models, and an independent cohort of 153 controls and individuals with schizophrenia from the FBIRN consortium to test them ( 41 ). Leveraging a graph-convolutional network for sFNC and a bidirectional LSTM for dFNC, we generated wide-brain and sub-network BAGs and evaluate their links to attention vigilance and working memory performance while controlling for demographic and clinical covariates. We hypothesize that ( 1 ) BAGs derived from dFNC will predict age as accurately as sFNC models, and ( 2 ) larger BAGs—particularly those based on dFNC—will associate with poorer cognitive performance, reflecting accelerated functional brain ageing in schizophrenia. By integrating dynamic connectivity with brain age modeling, this study provides the first direct evidence linking aberrant network dynamics to cognitive impairment and establishes dFNC-based BAG as a clinically relevant biomarker. 2. METHODS 2.1 Study Population Our study leveraged multiple datasets, including the UK Biobank or UKBB ( 36 , 37 ) and the Human Connectome Project’s Young Adult or HCP-YA ( 38 ) and Aging or HCP-A ( 40 ) cohorts to develop brain age prediction models. For the UKBB dataset, exclusion criteria encompassed a range of mental and behavioral disorders as categorized by the International Classification of Diseases version 10 (ICD-10): delirium not induced by alcohol and other psychoactive substances (F05); mental disorders due to brain damage, dysfunction, or physical disease (F06); personality and behavioral disorders due to brain disease, damage, or dysfunction (F07); unspecified organic or symptomatic mental disorders (F09); disorders due to psychoactive substance use (F10-F19); schizophrenia, schizotypal, and delusional disorders (F20-29); manic episodes (F30); and bipolar affective disorder (F31). We also excluded individuals who had sought treatment for “nerves, anxiety, tension, or depression” from either their general practitioner or a psychiatrist. After exclusions, the UKBB dataset comprised 15,978 individuals aged 45 to 82 years, with a mean age of 64.26 ± 7.55 years and a median age of 65 years, including 7,244 females and 8,734 males ( 36 , 37 ). All UKBB participants provide informed consent as part of the ethical oversight maintained by a dedicated Ethics and Guidance Council, which collaborates with UKBB to uphold an Ethics and Governance Framework. Additionally, the study received IRB approval from the Northwest Multicenter Research Ethics Committee. The HCP-YA dataset included 833 individuals aged 22 to 35 years, with a mean age of 28.66 ± 3.66 years and a median age of 29 years, consisting of 443 females and 390 males ( 39 ). The HCP-A dataset included 711 individuals aged 36 to 100 years, with a mean age of 60.54 ± 15.68 years and a median age of 58.67 years, consisting of 403 females and 308 males ( 40 ). Collectively, the datasets encompassed 17,522 individuals aged 22 to 100 years, with a mean age of 62.42 ± 10.96 years and a median age of 64 years, including 8,090 females and 9,432 males. All subject recruitment procedures and informed consent forms, including consent to share de-identified data, were approved by the Washington University IRB. Notably, the HCP datasets provided around 15 minutes of resting-state fMRI data per session, while the UKBB dataset included around 6 minutes of resting-state fMRI data. Because these data set contain multiple scans per participant, there are a total of 22,569 distinct resting-state fMRI scans in the training set. Table 1 shows the demographic information of both training and test dataset. View this table: View inline View popup Download powerpoint Table 1. The Training Data set is composed of UKBB, HCP, and HCP-A datasets, while the FBIRN data set composes the Test Data. Additionally, data from the Functional Imaging Biomedical Informatics Research Network (FBIRN) were employed to estimate the brain age gap (BAG) and examine its link with cognitive metrics such as working memory and attention ( 41 ). For the FBIRN dataset, raw imaging data were collected across seven sites: University of California, Irvine; University of California, Los Angeles; University of California, San Francisco; Duke University/University of North Carolina at Chapel Hill; the University of Iowa; the University of Minnesota; and the University of New Mexico. Each participant provided written informed consent, with protocols approved by the institutional review boards at each respective site. All subjects with schizophrenia (SZ) were clinically stable at the time of scanning. Diagnoses were validated through the Structured Clinical Interview for DSM-IV (SCID-IV), while healthy controls (HC) were assessed using the SCID-I/NP to confirm the absence of schizophrenia. Exclusion criteria for HC included a current or history of major neurological or psychiatric disorders and having a first-degree relative with an Axis-I psychotic disorder, as determined by SCID evaluations. The FBIRN dataset contains individuals between 18 and 62 years old, with a mean age of 37.88 and a median age of 38. Out of 311 individuals in the set, 151 of them have a diagnosis of schizophrenia; these participants are between 18 and 62 years old, with a mean age of 38.77 and a median age of 39. Overall, there were 81 female participants and 230 male participants; among those diagnosed with schizophrenia, there were 36 female participants and 115 male participants. For the FBIRN dataset, written informed consent was obtained from all participants. Institutional review boards approved the consent process of each study site. Table 1 presents the demographic information for all FBIRN participants, while Supplementary Table 1 details the demographic and clinical data for each site separately. Figure 1A illustrates the count density of the training datasets (UKBB, HCP-YA, and HCP-A), while Figure 1B displays the count density of the test dataset (FBIRN) used in our study. Download figure Open in new tab Download figure Open in new tab Download figure Open in new tab Download figure Open in new tab Figure 1. A) Count density of the training dataset used in the brain age prediction model employing both sFNC and dFNC variables. The training set comprises 17,522 individuals aged 22 to 100 years, including 8,090 females and 9,432 males. B) Count density of the FBIRN dataset used as test data to examine the link between BAG and cognitive/attention measures. The dataset includes 311 individuals aged 18 to 62 years (81 females and 230 males); our analysis focused on those with schizophrenia. C) NeuroMark pipeline extraction of 53 regions across seven networks: subcortical (SCN), auditory (AUD), sensorimotor (SMN), visual (VSN), cognitive control (CCN), default mode (DMN), and cerebellar (CBN). Both static and dynamic functional network connectivity (FNC) were computed between each pair of regions using the NeuroMark template, yielding a 53×53 connectivity matrix. A representative mean FNC was estimated from UK Biobank participants. D) Illustration of sFNC data processing: Training data (UKBB, HCP, and HCP-Aging) are input into an untrained model with chronological age as the label. The trained model then predicts brain age from test data (FBIRN). Comparison of predicted versus chronological age determines the brain age gap. E) Illustration of the dFNC data processing, analogous to panel D, demonstrating the training and application of the model using dynamic connectivity data. 2.2 Imaging Protocol For UKBB, the imaging data were collected using Siemens Skyra 3T scanners. The resolution was 2.4 × 2.4 × 2.4 mm 3 . The data was collected as 490 timeframes over 6 minutes, with a TR of 0.735 s, a TE of 39ms, and a flip angle of 52°. During the scan, participants are instructed to fixate on a cross displayed on a screen to minimize eye movement and maximize data consistency. For HCP, the imaging data were collected using a Siemens 3T scanner as 1200 frames per run over 14:33 minutes per run, with 4 runs total. The TR was 720 ms, the TE was 33.1, the flip angle was 52°, and the slice thickness was 2.0mm. For the HCP-YA, imaging data were captured using a 3T Siemens Prisma scanner equipped with a 32-channel phased-array head coil. The rs-fMRI acquisition parameters included a TR of 720 ms, a TE of 33.1 ms, and a field of view of 208 × 180 mm². A flip angle of 52° was used, with images obtained across 72 oblique-axial slices at a resolution of 2.0 × 2.0 × 2.0 mm³. Each scanning session was conducted over a duration of approximately 14 minutes and 40 seconds. For the HCP-A imaging data were collected using a 3T Siemens Prisma scanner equipped with a 32-channel phased-array head coil. The rs-fMRI protocol included the following parameters: TR of 800 ms, TE of 37 ms, a field of view of 810 × 936 mm², and a flip angle of 52°. Images were acquired with a resolution of 2.0 × 2.0 × 2.0 mm³ across 72 oblique-axial slices. Each scanning session lasted approximately 14 minutes and 40 seconds. Similar to the UKBB, participants in the HCP are asked to fixate on a cross presented on a screen during the scan to reduce eye movements and ensure consistent data collection. For FBIRN, imaging data were acquired using six Siemens 3T scanners and one General Electric 3T scanner. All sites followed a uniform rs-fMRI protocol. T2*-weighted functional images were captured using an echo-planar imaging sequence, aligned along the anterior and posterior commissure (AC-PC) line, with the following parameters: TE of 30 ms, TR of 2 s, flip angle of 77°, slice gap of 1 mm, voxel dimensions of 3.4 × 3.4 × 4 mm³, and a series of 162 frames spanning 5 minutes and 38 seconds. During scans, participants were instructed to keep their eyes closed. 2.3 Data Processing Data from fMRI were preprocessed using Statistical Parametric Mapping (SPM12, https://www.fil.ion.ucl.ac.uk/spm/ ) in the MATLAB 2019 environment. We conducted a rigid body motion correction with SPM’s toolbox to address head motion. Subsequently, the imaging data were spatially normalized to an echo-planar imaging (EPI) template in standard Montreal Neurological Institute (MNI) space. Finally, a Gaussian kernel with a full width at half maximum (FWHM) of 6 mm was applied to smooth the fMRI images. 2.4 Extracting independent components using NeuroMark To obtain reliable independent components (ICs), we employed the NeuroMark fully automated ICA pipeline, which integrates previously derived component maps as spatial constraints. The NeuroMark framework utilizes templates developed from substantial datasets, specifically the Human Connectome Project (HCP: https://www.humanconnectome.org/study/hcp-young-adult/document/1200-subjects-data-release , 823 subjects after the subject selection) and the Genomics Superstruct Project (GSP: https://dataverse.harvard.edu/dataverse/GSP , 1005 subjects post-selection). This approach has proven effective across numerous studies, identifying a broad array of imaging markers for various brain disorders. Further information on template development is available in our prior publication on the NeuroMark method ( 42 ). The NeuroMark template includes 53 independent components (ICs), categorized into seven functional networks: subcortical (SCN), auditory (AUD), sensorimotor (SMN), visual (VSN), cognitive control (CCN), default-mode (DMN), and cerebellar networks (CBN), as shown in Figure 1C and Supplementary Figure 1 . Supplementary Table 2 also shows all 53 ICs and their interactions. For the ICA analysis, we utilized these templates via the NeuroMark_fMRI_1.0 template, accessible through GIFT v4.0.5.14 GIFT ( http://trendscenter.org/software/gift and on the TReNDS website @ http://trendscenter.org/data ). Additional denoising and artifact removal steps prior to calculating dynamic functional connectivity included: 1) linear, quadratic, and cubic de-trending; 2) multiple regression of the six realignment parameters and their temporal derivatives; 3) outlier removal; and 4) low pass filtering below a frequency of 0.15 Hz. 2.6 Static functional network connectivity To estimate sFNC, we calculated the Pearson correlation between pairs of ICs in each subject as shown in equation 1 where x 1 and x 2 are time course signals and and are the mean of x 1 and x 2, respectively. This Pearson correlation takes values in the interval [− 1, 1] and measures the strength of the linear relationship between x 1 and x 2. Each FNC is a 53×53 matrix, from which we derived a total of connectivity features. Figure 1C shows a representative FNC we used in our study. 2.7 Dynamic functional network connectivity To compute dynamic functional network connectivity (dFNC), we employed a sliding window approach. We used a tapered window, formed by convolving a rectangular window with the duration of 20TR (UKBB: 14.7 s, HCP-YA: 14.4 s, HCP-A: 16 s, FBIRN: 40s) with a Gaussian kernel (σ = 3), to precisely focus on data at each time point (see Supplementary Figure 2 ) ( 30 , 31 , 33 , 34 , 48 , 50 , 52 ). This approach, detailed in Equation 1 , was utilized to compute the FNC at each time point. We then aggregated the dFNC estimates for each window and each subject to construct a three-dimensional array (C × C × T), where C represents the 53 independent components, and T represents the number of windows. This array captures the temporal variations in connectivity among the independent components. 2.8 Training the model In our study, the resting-state fMRI time series were uniformly downsampled to the shortest duration across datasets, corresponding to approximately 5.4 minutes. We then computed dFNC matrices using a sliding window approach with a window size of 40 TRs. To evaluate model stability, K=5-fold cross-validation was performed on the training set of 22,569 scans from the UKB, HCP-YA and HCP-A data sets. For sFNC, we deployed a brain connectivity graph convolutional network or BCGCN to predict the brain age ( 43 , 44 ). The BCGCN model utilized ReLU activation for all layers except the output, which employed a linear activation (see Supplementary Figure 3A ). For the analysis of the dFNC data, we employed a bi-directional Long Short-Term Memory network (biLSTM) ( 45 ) configured with three recurrent layers with 128 hidden units each, a dropout rate of 0.1, and a fully connected regression layer to predict brain age (see Supplementary Figure 3B ). In both models, we utilized an Adam optimizer targeting Mean Absolute Error, with a learning rate of 1e -3 and a batch size of 64, conducting training over 100 epochs. The epoch with the optimal cross-validation performance was selected for inference on the test set. Figure 1D and Figure 1E illustrate the modeling and testing procedures for the sFNC and dFNC features, respectively. 2.9 Statistical analysis We developed brain age prediction models incorporating both sFNC and dFNC variables. These models were then applied to the FBIRN dataset to estimate the predicted brain age and subsequently brain age gap (BAG) for each participant. To investigate the associations between these BAGs and cognitive measures, including working memory and attention vigilance, we constructed a General Linear Model (GLM). This model included age, sex, site, age 2 , the interaction of age and sex, and the diagnosis (schizophrenia or control) as covariates. For each cognitive measure, we tested two hypotheses—assessing the association of BAG derived from two modalities (sFNC-based and dFNC-based BAGs). We corrected for multiple comparisons across the two modalities using the False Discovery Rate (FDR) correction method to ensure the robustness of our findings ( 46 ). 3. RESULTS 3.1 Training and validating the brain age prediction model on the healthy population After developing brain age prediction models using wide-brain static and dynamic functional network connectivity (sFNC and dFNC, respectively), we validated their performance on a validation dataset within each run. This step was essential to assess the models’ generalizability and accuracy in predicting brain age. We calculated the Pearson correlation coefficients between the chronological brain age and the predicted brain age to evaluate the models’ performance. Results demonstrated a strong correlation between the predicted brain age from the sFNC-based model and chronological age, with a correlation value of 0.8755 ( Figure 2A ). Similarly, the dFNC-based model also showed a strong correlation, with a correlation value of 0.8675, illustrating its effectiveness in age prediction ( Figure 2B ). Download figure Open in new tab Figure 2. Correlation between brain age predictions and chronological age in sFNC- and dFNC-based models. These graphs illustrate the relationship between chronological age and predicted brain age in the training dataset (UK Biobank, HCP, and HCP Aging). A) sFNC-based predictions (green) show a strong positive correlation with chronological age (r = 0.8324, p < 1.1755e -38 ). B) dFNC-based predictions (orange) also show a significant positive correlation (r = 0.801, p < 1.1755e -38 ). 3.2 Static and dynamic FNC-based wide-brain age gap links with attention vigilance After developing and validating the brain age prediction models, we applied them to the FBIRN dataset to calculate wide-brain BAGs (wBAGs), representing brain age gaps estimated from wide-brain sFNC and dFNC for each individual. We then constructed GLMs to examine the association between wBAGs and attention vigilance while adjusting for age, sex, study site, age 2 , the age-by-sex interaction, and diagnosis as covariates. The corresponding p-values were further adjusted to account for testing two modalities. Because our age prediction model performs less accurately for individuals under 38 years old (see Figure 2 ), we restricted the FBIRN analysis to participants over 38. This resulted in a final sample of 153 participants (121 males) with a mean age of 47.73□±5.94 years. Figure 3A shows a significant negative association between sFNC-based wBAGs and attention vigilance ( r □=-0.2923, β □= −0.9860, SE =0.2829, 95% CI : −1.5454 to −0.4263, FDR p =0.0013, N =153). Figure 3B demonstrates a significant association between dFNC-based wBAGs and attention vigilance ( r □=□-0.2715, β =-0.4395, S □=0.1366, 95% CI : −0.7099 to −0.1691, FDR p =0.0016, N □=□153). In both models, we observed a negative association between wBAGs and cognitive performance, indicating that individuals with older-appearing brains tend to perform worse on attention vigilance tasks. Download figure Open in new tab Figure 3. Associations between sFNC- and dFNC-based wide-brain age gap (wBAG) and cognitive performance in individuals with schizophrenia. After developing and validating the brain age prediction models, we applied them to the FBIRN dataset to calculate wBAGs based on static (sFNC) and dynamic (dFNC) functional network connectivity features. Analyses were restricted to individuals older than 38 years due to reduced model accuracy in younger participants (see Figure 2 ), resulting in a final sample of 160 participants (mean age = 47.73□±□5.94 years; 121 males). General Linear Models (GLMs) were used to examine associations between wBAGs and two cognitive domains—attention vigilance and working memory—controlling for age, sex, study site, age squared, age-by-sex interaction, and diagnosis. P-values were adjusted for multiple comparisons across modalities using the Benjamini-Hochberg FDR correction. A) sFNC-based wBAG was significantly negatively associated with attention vigilance ( r =–0.2923, β =–0.9860, SE =0.2829, 95% CI : –1.5454 to – 0.4263, FDR p =0.0013, N =153). B) dFNC-based wBAG was significantly negatively associated with attention vigilance ( r =–0.2715, β =–0.4395, SE =0.1366, 95% CI : –0.7099 to –0.1691, FDR p =0.0016, N=153). C) sFNC-based wBAG was significantly negatively associated with working memory ( r =–0.2237, β =–0.9603, SE =0.3615, 95% CI : –1.6761 to –0.2445, FDR p =0.0088, N =153). D) dFNC-based wBAG was significantly negatively associated with working memory ( r =–0.2508, β =–0.5195, SE =0.1732, 95% CI : –0.8621 to –0.1769, FDR p =0.0064, N =153). Across both cognitive domains, higher wBAGs—reflecting accelerated brain aging—were associated with poorer cognitive performance. 3.3 Static and dynamic FNC-based wide-brain age gap links with working memory We constructed GLMs to examine the association between sFNC- and dFNC-based wBAGs and working memory while adjusting for age, sex, study site, age 2 , age-by-sex interaction, and diagnosis as covariates. P-values were further corrected for multiple comparisons across the two modalities. Figure 3C shows a significant negative association between sFNC-based wBAGs and working memory ( r =-0.2237, β =-0.9603, SE =0.3615, 95% CI =-0.2918 to −0.0856, FDR p =0.0088, N =153). Figure 3D demonstrates a significant negative association between dFNC-based wBAGs and working memory ( r =-0.2508, β =-0.5195, SE =0.1732, 95% CI = −0.8621 to −0.1769, FDR p =0.0064, N =151). These negative correlations indicate that individuals with higher brain age relative to chronological age tend to exhibit poorer working memory performance. 3.4 Sub-network brain age gaps predict attention vigilance and working memory The next question was whether sFNC- and dFNC-based sub-network brain age gaps or subBAGs predict attention vigilance, and working memory performance. To address this, we developed brain age prediction models separately for each network using data from our healthy population. Specifically, we created seven distinct brain age prediction models corresponding to seven networks including SCN, AUC, SMN, VSN, CCN, DMN, and CBN. The correlation between chronological age and predicted brain age for each sub-network is presented in Supplementary Figure 4 . Next, we examined the association between subBAGs and attention vigilance, and working memory, controlling for age, sex, site, age², age-by-sex interaction, and diagnosis as covariates. Given seven networks and two modalities (sFNC and dFNC), a total of 14 statistical tests were performed, and p-values were corrected accordingly. Figure 4.A illustrates the relationship between sFNC- and dFNC subBAGs and attention vigilance. Each bar represents the association between subBAGs and attention vigilance, with crosshatched bars indicating dFNC-based subBAGs and solid bars representing sFNC-based subBAGs. Bars showing a significant association after FDR correction are marked with double asterisks. As shown, the sFNC-based subBAGs estimated from the subcortical network exhibit a negative association with attention vigilance ( r = −0.2357, β = −0.1813, SE = 0.0655, 95% CI : - 0.3111 to −0.0516, FDR p = 0.0228, N = 153). Similarly, the dFNC-based subBAGs from the subcortical network show a significant negative association with attention vigilance ( r =-0.2806, β =-0.2989, SE =0.0896, 95% CI : −0.4763 to −0.1215, FDR p =0.0140, N =153). Additionally, dFNC-based subBAGs estimated from the sensorimotor network show a significant negative association ( r = −0.2384, β =-0.2153, SE =0.0769, 95% CI : −0.3675 to −0.0631, FDR p =0.0228, N =153) and from the default mode network ( r =-0.2666, β =-0.1053, SE =0.0333, 95% CI : - 0.1713 to −0.0392, FDR p =0.0140, N =153). Our results also indicate a significant negative association between sFNC-based subBAGs estimated from the cognitive control network and attention vigilance ( r =-0.2081, β =-0.0353, SE =0.0145, 95% CI : −0.0641 to −0.0065, FDR p =0.0455, N =153). Supplementary Table 3 presents the complete set of models examining the association between subBAGs and attention vigilance. Download figure Open in new tab Download figure Open in new tab Figure 4. Associations between subBAGs and cognitive performance. A) Relationship between subBAGs and attention vigilance. Crosshatched bars represent dFNC -based subBAGs and solid bars represent sFNC - based subBAGs. Bars with double asterisks indicate significant associations after FDR correction. Significant negative associations were observed for sFNC - based subBAGs from the subcortical network ( r =-0.2357, β =-0.1813, SE =0.0655, 95% CI =-0.3111 to −0.0516, FDR p =0.0228, N =153) and dFNC - based subBAGs from the subcortical network ( r =-0.2806, β =-0.2989, SE =0.0896, 95% CI =-0.4763 to −0.1215, FDR p =0.0140, N =153). Additional significant negative associations were observed for dFNC -based subBAGs from the sensorimotor network ( r =-0.2384, β =-0.2153, SE =0.0769, 95% CI =-0.3675 to −0.0631, FDR p =0.0228, N =153) and the default mode network ( r =-0.2666, β =-0.1053, SE =0.0333, 95% CI =-0.1713 to −0.0392, FDR p =0.0140, N =153). The sFNC-based subBAGs from the cognitive control network also showed a significant negative association ( r =-0.2081, β =-0.0353, SE =0.0145, 95% CI =-0.0641 to −0.0065, FDR p =0.0455, N =153). B) Relationship between subBAGs and working memory. Significant negative associations ( uncorrected p <0.05) were identified for sFNC - based subBAGs from the subcortical network ( r =-0.1724, β =-0.1050, SE =0.0518, 95% CI =-0.2076 to −0.0020, uncorrected p =0.0447, N =153), dFNC -based subBAGs from the subcortical network ( r =-0.1754, β =-0.1473, SE =0.0714, 95% CI =-0.2885 to −0.0060, uncorrected p =0.0410, N =153), and dFNC-based subBAGs from the default mode network ( r =-0.2114, β =-0.0662, SE =0.0264, 95% CI =-0.1186 to −0.0139, uncorrected p =0.0134, N =153). However, none of these associations survived FDR correction. All models were adjusted for age, sex, site, age², age-by-sex interaction, and diagnosis. Figure 4.B shows the link between sFNC- and dFNC-based subBAGs with working memory using GLM with age, sex, site, age², age-by-sex interaction, and diagnosis as covariates. Among all models, we identified only three showing a significant association between subBAGs and working memory (as shown with single asterisk): sFNC-based subBAGs estimated from the subcortical network ( r =-0.1724, β = −0.1050, SE = 0.0518, 95% CI : −0.2076 to −0.0020, uncorrected p =0.0447, N =153), dFNC-based subBAGs estimated from the subcortical network ( r =-0.1754, β =-0.1473, SE =0.0714, 95% CI : −0.2885 to −0.0060, uncorrected p =0.0410, N =153), and dFNC-based subBAGs estimated from the default mode network ( r =-0.2114, β =-0.0662, SE =0.0264, 95% CI : −0.1186 to −0.0139, uncorrected p =0.0134, N =153). However, none of these associations remained significant after FDR correction. Supplementary Table 4 presents the complete set of models examining the association between subBAGs and working memory. DISCUSSION The brain-age gapD(BAG), the difference between a person’s predicted brain age and chronological age, has emerged as a compact indicator of neurobiological health: a larger positive BAG signals accelerated ageing, whereas a negative BAG reflects resilience. Schizophrenia reliably shows an elevated BAG, and larger BAGs correlate with poorer working-memory performance□ ( 47 , 49 , 51 ). Crucially, almost all BAG work to date relies on structural MRI or static FNC, overlooking the millisecond-scale reconfiguration of brain networks. We present the first dynamic-FNC brain age models—at both wide-brain (wBAG) and network-specific (subBAG) levels—leveraging time-resolved connectivity to capture the neural volatility that likely underlies cognitive dysfunction. By testing dFNC-BAG against working memory and attention vigilance performance in schizophrenia cohort, we aim to uncover a previously inaccessible layer of pathophysiology and position dFNC-BAG as new precision biomarker capable of guiding circuit-based interventions and accelerating translational psychiatry. Extant research shows that sustained attention declines with normal aging ( 53 ), and schizophrenia magnifies this decline, with patients repeatedly scoring well below normative thresholds across sites, ages, and sexes□ ( 54 ). Our data reveal a robust negative association between attention vigilance measure and both wBAG and subBAG indices of brain age acceleration—whether derived from static or dynamic FNC. Put simply, the “older” the brain looks, the poorer the patient performs. This powerful link spotlights accelerated macro- and micro-scale brain aging as a driving force behind attentional breakdown in schizophrenia and elevates BAG to a precision and network-level biomarker with clear therapeutic implications. Convergent neuroimaging work has already mapped attentional control to frontoparietal and default mode circuitry; our findings now tie those circuits’ temporal dysmaturation directly to real-world cognitive deficits. Neuroimaging studies have localized specific brain regions and connectivity networks that facilitate attentional processes. For instance, processing temporal stimuli has been linked to activity in the pre-supplementary motor area ( 55 ), whereas joint attention engages the ventromedial frontal cortex, cingulate cortex, caudate nuclei, and left superior frontal gyrus ( 56 ). Mirroring these maps, our strongest BAG-attention associations localize to sensorimotor and subcortical networks—notably the caudate, thalamus, and putamen—underscoring their central role in the disorder’s cognitive burden. Overall, this work bridges neuroanatomical change with clinical symptomatology and supports BAG as a precision tool for early identification and stratification in neuropsychiatric disorders. Working memory deficits—spanning visuospatial, phonological, and executive domains—are a core cognitive feature of schizophrenia, consistently observed across cohorts relative to healthy controls ( 57 ). These deficits are compounded by normative age-related decline, with working memory performance deteriorating progressively across adulthood ( 58 ). Building on this foundation, we examined the relationship between BAG and working memory using both sFNC and dFNC models. Both sFNC- and dFNC-based wBAG showed robust negative associations with working memory performance in the FBIRN cohort. In other words, greater brain age acceleration predicted worse working memory capacity. This finding extends prior work by linking functional BAG to cognition in schizophrenia, reinforcing the relevance of BAG as a mechanistic marker of cognitive vulnerability. Notably, our results highlight the subcortical and default mode networks as key contributors to this relationship, pointing to specific neural systems where accelerated ageing most profoundly impacts working memory function. Notably, the association between wBAGs and working memory, and subBAGs and attention vigilance, was stronger for dFNC-derived BAG, suggesting cognitive deficits in schizophrenia may be more closely linked to dynamic reconfiguration of brain networks than to static connectivity patterns. This aligns with evidence that cognition depends on flexible, moment-to-moment network adaptations to meet cognitive demands ( 59 ). By capturing temporal fluctuations in connectivity, the dFNC approach provides a richer, more ecologically valid representation of brain function, making it more sensitive to the cognitive disruptions observed in schizophrenia. In contrast, sFNC averages connectivity over time, potentially obscuring critical dynamic alterations. The higher temporal resolution of dFNC also yields more data points per scan, enhancing the detection of subtle brain–behavior associations ( 30 , 31 , 33 ). These findings highlight the unique potential of dFNC-based BAG as a precision biomarker to capture clinically meaningful cognitive decline and inform individualized therapeutic strategies in schizophrenia. Our study has several limitations that merit consideration. Firstly, the choice of window size in dynamic connectivity analysis inherently assumes certain characteristics about temporal dynamics. Shorter windows capture rapid fluctuations effectively, while longer windows tend to smooth these fluctuations, potentially obscuring meaningful variability. Future research should explore a broader range of window sizes to more comprehensively assess their impact on brain connectivity measures ( 60 ). Additionally, our study excluded participants with any health conditions from the UKBB dataset. Future studies may benefit from including a broader cohort by aggregating data from multiple datasets that include healthy individuals, which could enhance the generalizability and robustness of the findings. Moreover, prior research suggests that factors like increased income or engagement in cognitively stimulating activities can mitigate age-related declines in working memory. However, our model did not account for variables such as average activity levels, diet, or other environmental factors that could also influence cognitive and brain function( 62 –64). Adjusting both brain and cognitive measures for these variables in the future studies could enhance our understanding of the complex interactions between lifestyle, health conditions, and cognitive aging, providing a more nuanced perspective on these relationships. Finally, due to deviations observed in our brain-age predictions for younger adults, likely driven by the limited representation of this age group in our healthy training dataset, we excluded FBIRN participants under 38. Expanding brain-age modeling efforts to include more younger adults from diverse datasets is an important next step, enabling exploration of whether BAG meaningfully captures cognitive deficits earlier in life and across the full adult lifespan. In conclusion, our study establishes BAG, derived from both static and dynamic functional connectivity at wide-brain and subnetwork levels, as a powerful and clinically relevant marker of cognitive deficits in schizophrenia. We demonstrate that higher BAG—reflecting accelerated brain aging—is significantly associated with worse working memory and attention vigilance, capturing core cognitive dysfunctions of the disorder. Importantly, the dynamic approach provides unique sensitivity by incorporating temporal fluctuations in connectivity, offering deeper insight into the brain’s functional organization than traditional static methods. Our findings highlight dFNC-based BAG as especially informative, uncovering stronger links to the neural substrates of cognitive decline. These results not only advance mechanistic understanding but also open new avenues for clinical translation. Future research expanding dynamic modeling and integrating lifestyle and health factors holds tremendous promise for optimizing prediction models. Overall, this work lays the foundation for next-generation precision psychiatry tools capable of early detection and personalized intervention strategies to mitigate cognitive deterioration and improve long-term outcomes in schizophrenia. Data Availability All data produced in the present study are available upon reasonable request to the authors Code and data availability statement The code used for preprocessing and FNC calculation are available at https://trendscenter.org/software/ . Also, statistical parametric mapping (SPM 12) is available at https://www.fil.ion.ucl.ac.uk/spm/ . The Neuromark framework and the Neuromark template (Neuromark_fMRI_1.0) have been made available and incorporated into the Group ICA Toolbox ( https://github.com/trendscenter/gift ). Users worldwide can now directly download and utilize these resources. We also use this https://www.nitrc.org/projects/bnv/ for brain graph. Competing interests Dr. Sendi has served as a consultant for Niji Corp for unrelated work. Dr. Mathalon has served as a consultant for Aptinyx, Boehringer-Ingelheim Pharmaceuticals, Cadent Therapeutics, and Greenwich Biosciences for unrelated work. The remaining authors declare no competing interests. Acknowledgements We thank the participants of this study and those involved in data collection. This work was supported by the National Institutes of Health (NIH) grants R01MH123610 and T32MH125786, the National Science Foundation (NSF) grant 2112455, and the Phyllis and Jerome Lyle Rappaport Mental Health Research Scholars Award. REFERENCES 1. ↵ Desai PR , Lawson KA , Barner JC , Rascati KL ( 2013 ): Estimating the direct and indirect costs for community-dwelling patients with schizophrenia . Journal of Pharmaceutical Health Services Research 4 : 187 – 194 . OpenUrl CrossRef 2. ↵ Kessler RC , Birnbaum H , Demler O , Falloon IRH , Gagnon E , Guyer M , et al. ( 2005 ): The prevalence and correlates of nonaffective psychosis in the National Comorbidity Survey Replication (NCS-R) . Biol Psychiatry 58 : 668 – 676 . OpenUrl CrossRef PubMed Web of Science 3. ↵ Guo JY , Ragland JD , Carter CS ( 2019 , May 1): Memory and cognition in schizophrenia . Molecular Psychiatry , vol. 24 . Nature Publishing Group , pp 633 – 642 . OpenUrl CrossRef PubMed 4. ↵ Van Snellenberg JX , Girgis RR , Horga G , van de Giessen E , Slifstein M , Ojeil N , et al. ( 2016 ): Mechanisms of Working Memory Impairment in Schizophrenia . Biol Psychiatry 80 : 617 – 626 . OpenUrl CrossRef PubMed 5. ↵ Dutterer J , Bansal S , Robinson B , Gold JM ( 2023 ) : Sustained attention deficits in schizophrenia: Effect of memory load on the Identical Pairs Continuous Performance Test . Schizophr Res Cogn 33 . doi: 10.1016/j.scog.2023.100288 OpenUrl CrossRef 6. ↵ Li YT , Zhang C , Han JC , Shang YX , Chen ZH , Cui G Bin , Wang W ( 2024 , January 1): Neuroimaging features of cognitive impairments in schizophrenia and major depressive disorder . Therapeutic Advances in Psychopharmacology , vol. 14 . SAGE Publications Ltd . doi: 10.1177/20451253241243290 OpenUrl CrossRef 7. ↵ Brent BK , Thermenos HW , Keshavan MS , Seidman LJ ( 2013 , October ): Gray Matter Alterations in Schizophrenia High-Risk Youth and Early-Onset Schizophrenia. A Review of Structural MRI Findings . Child and Adolescent Psychiatric Clinics of North America , vol. 22 . pp 689 – 714 . OpenUrl CrossRef PubMed 8. ↵ Penadés R , Franck N , González-Vallespí L , Dekerle M ( 2019 ): Neuroimaging Studies of Cognitive Function in Schizophrenia . Advances in Experimental Medicine and Biology , vol. 1118 . Springer New York LLC , pp 117 – 134 . OpenUrl PubMed 9. ↵ Gur RE , Gur RC ( 2010 ): Functional magnetic resonance imaging in schizophrenia . Dialogues Clin Neurosci 12 : 333 – 343 . OpenUrl PubMed 10. ↵ Wrigglesworth J , Yaacob N , Ward P , Woods RL , McNeil J , Storey E , et al. ( 2022 ): Brain-predicted age difference is associated with cognitive processing in later-life . Neurobiol Aging 109 : 195 – 203 . OpenUrl CrossRef PubMed 11. Anatürk M , Kaufmann T , Cole JH , Suri S , Griffanti L , Zsoldos E , et al. ( 2021 ) : Prediction of brain age and cognitive age: Quantifying brain and cognitive maintenance in aging . Hum Brain Mapp 42 : 1626 – 1640 . OpenUrl CrossRef PubMed 12. ↵ Elliott ML , Belsky DW , Knodt AR , Ireland D , Melzer TR , Poulton R , et al. ( 2021 ): Brain-age in midlife is associated with accelerated biological aging and cognitive decline in a longitudinal birth cohort . Mol Psychiatry 26 : 3829 – 3838 . OpenUrl CrossRef PubMed 13. ↵ Stefaniak JD , Mak E , Su L , Carter SF , Dounavi M-E , Muniz Terrera G , et al. ( 2024 ): Brain age gap, dementia risk factors and cognition in middle age . Brain Commun 6 . doi: 10.1093/braincomms/fcae392 OpenUrl CrossRef 14. ↵ Joo Y , Namgung E , Jeong H , Kang I , Kim J , Oh S , et al. ( 2023 ): Brain age prediction using combined deep convolutional neural network and multi-layer perceptron algorithms . Sci Rep 13 . doi: 10.1038/s41598-023-49514-2 OpenUrl CrossRef 15. ↵ Shahab S , Mulsant BH , Levesque ML , Calarco N , Nazeri A , Wheeler AL , et al. ( 2019 ): Brain structure, cognition, and brain age in schizophrenia, bipolar disorder, and healthy controls . Neuropsychopharmacology 44 : 898 – 906 . OpenUrl CrossRef PubMed 16. ↵ Koutsouleris N , Davatzikos C , Borgwardt S , Gaser C , Bottlender R , Frodl T , et al. ( 2014 ) : Accelerated brain aging in schizophrenia and beyond: A neuroanatomical marker of psychiatric disorders . Schizophr Bull 40 : 1140 – 1153 . OpenUrl CrossRef PubMed 17. ↵ Schnack HG , Van Haren NEM , Nieuwenhuis M , Pol HEH , Cahn W , Kahn RS ( 2016 ) : Accelerated brain aging in schizophrenia: A longitudinal pattern recognition study . American Journal of Psychiatry 173 : 607 – 616 . OpenUrl CrossRef PubMed 18. ↵ Hajek T , Franke K , Kolenic M , Capkova J , Matejka M , Propper L , et al. ( 2019 ): Brain Age in Early Stages of Bipolar Disorders or Schizophrenia . Schizophr Bull 45 : 191 – 198 . OpenUrl CrossRef 19. ↵ Nenadić I , Dietzek M , Langbein K , Sauer H , Gaser C ( 2017 ): BrainAGE score indicates accelerated brain aging in schizophrenia, but not bipolar disorder . Psychiatry Res Neuroimaging 266 : 86 – 89 . OpenUrl CrossRef 20. ↵ Wang J , Kochunov P , Sampath H , Hatch KS , Ryan MC , Xue F , et al. ( 2021 ): White matter brain aging in relationship to schizophrenia and its cognitive deficit . Schizophr Res 230 : 9 – 16 . OpenUrl CrossRef PubMed 21. ↵ Zhu JD , Wu YF , Tsai SJ , Lin CP , Yang AC ( 2023 ): Investigating brain aging trajectory deviations in different brain regions of individuals with schizophrenia using multimodal magnetic resonance imaging and brain-age prediction: a multicenter study . Transl Psychiatry 13 . doi: 10.1038/s41398-023-02379-5 OpenUrl CrossRef PubMed 22. Millar PR , Luckett PH , Gordon BA , Benzinger TLS , Schindler SE , Fagan AM , et al. ( 2022 ): Predicting brain age from functional connectivity in symptomatic and preclinical Alzheimer disease . Neuroimage 256 . doi: 10.1016/j.neuroimage.2022.119228 OpenUrl CrossRef 23. Monti RP , Gibberd A , Roy S , Nunes M , Lorenz R , Leech R , et al. ( 2020 ): Interpretable brain age prediction using linear latent variable models of functional connectivity . PLoS One 15 . doi: 10.1371/journal.pone.0232296 OpenUrl CrossRef PubMed 24. ↵ Sendi MS , Salat DH , Calhoun VD ( 2021 ): Brain age gap difference between healthy and mild dementia subjects: Functional network connectivity analysis . 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society . Virtual Conference: IEEE . 25. ↵ Lund MJ , Alnæs D , de Lange AMG , Andreassen OA , Westlye LT , Kaufmann T ( 2022 ): Brain age prediction using fMRI network coupling in youths and associations with psychiatric symptoms . Neuroimage Clin 33 . doi: 10.1016/j.nicl.2021.102921 OpenUrl CrossRef PubMed 26. ↵ Chang JR , Yao ZF , Hsieh S , Nordling TEM ( 2024 ): Age Prediction Using Resting-State Functional MRI . Neuroinformatics 22 : 119 – 134 . OpenUrl PubMed 27. ↵ Sakoğlu Ü , Pearlson GD , Kiehl KA , Wang YM , Michael AM , Calhoun VD ( 2010 ) : A method for evaluating dynamic functional network connectivity and task-modulation: Application to schizophrenia . Magnetic Resonance Materials in Physics, Biology and Medicine 23 : 351 – 366 . OpenUrl CrossRef 28. ↵ Calhoun VD , Miller R , Pearlson G , Adali T ( 2014 ) : The Chronnectome: Time-Varying Connectivity Networks as the Next Frontier in fMRI Data Discovery . Neuron 84 : 262 – 274 . OpenUrl CrossRef PubMed Web of Science 29. ↵ Lurie DJ , Kessler D , Bassett DS , Betzel RF , Breakspear M , Keilholz S , et al. ( 2020 , February 1): Questions and controversies in the study of time-varying functional connectivity in resting fMRI . Network Neuroscience , vol. 4 . MIT Press Journals , pp 30 – 69 . OpenUrl CrossRef PubMed 30. ↵ Sendi MSE , Pearlson GD , Daniel , H JMF , Preda A , Erp TGM van , D.Calhoun V ( 2020 ): Multiple overlapping dynamic patterns of the visual sensory network in schizophrenia . Schizophr Res 228 : 103 – 111 . OpenUrl 31. ↵ Sendi MSE , Zendehrouh E , Ellis CA , Fu Z , Chen J , Miller RL , et al. ( 2023 ): The link between static and dynamic brain functional network connectivity and genetic risk of Alzheimer’s disease . Neuroimage Clin 37 : 103363 . OpenUrl PubMed 32. Sendi MS , Fu Z , Harnett N , van Rooij S , Vergara V , Pizzagalli D , et al. ( 2024 ): Brain dynamics reflecting an intra-network brain state is associated with increased posttraumatic stress symptoms in the early aftermath of trauma . Nature Mental Health . 33. ↵ Sendi MSE , Zendehrouh E , Miller RL , Fu Z , Du Y , Liu J , et al. ( 2021 ) : Alzheimer’s Disease Projection From Normal to Mild Dementia Reflected in Functional Network Connectivity: A Longitudinal Study . Front Neural Circuits 14 . 34. ↵ Dini H , Sendi MSE , Sui J , Fu Z , Espinoza R , Narr KL , et al. ( 2021 ): Dynamic functional connectivity predicts treatment response to electroconvulsive therapy in major depressive disorder . Front Hum Neurosci 15 : 1 – 11 . OpenUrl CrossRef PubMed 35. ↵ Allen EA , Damaraju E , Plis SM , Erhardt EB , Eichele T , Calhoun VD ( 2014 ): Tracking whole-brain connectivity dynamics in the resting state . Cerebral Cortex 24 : 663 – 676 . OpenUrl CrossRef PubMed Web of Science 36. ↵ Miller KL , Alfaro-Almagro F , Bangerter NK , Thomas DL , Yacoub E , Xu J , et al. ( 2016 ): Multimodal population brain imaging in the UK Biobank prospective epidemiological study . Nat Neurosci 19 : 1523 – 1536 . OpenUrl CrossRef PubMed 37. ↵ Littlejohns TJ , Holliday J , Gibson LM , Garratt S , Oesingmann N , Alfaro-Almagro F , et al. ( 2020 , December 1): The UK Biobank imaging enhancement of 100,000 participants: rationale, data collection, management and future directions . Nature Communications , vol. 11 . Nature Research . 38. ↵ Glasser MF , Smith SM , Marcus DS , Andersson JLR , Auerbach EJ , Behrens TEJ , et al. ( 2016 ): The Human Connectome Project’s neuroimaging approach . Nat Neurosci 19 : 1175 – 1187 . OpenUrl CrossRef PubMed 39. ↵ Van Essen DC , Smith SM , Barch DM , Behrens TEJ , Yacoub E , Ugurbil K ( 2013 ) : The WU-Minn Human Connectome Project: An overview . Neuroimage 80 : 62 – 79 . OpenUrl CrossRef PubMed Web of Science 40. ↵ Bookheimer SY , Salat DH , Terpstra M , Ances BM , Barch DM , Buckner RL , et al. ( 2019 ) : The Lifespan Human Connectome Project in Aging: An overview . Neuroimage 185 : 335 – 348 . OpenUrl CrossRef PubMed 41. ↵ Keator DB , van Erp TGM , Turner JA , Glover GH , Mueller BA , Liu TT , et al. ( 2016 ): The Function Biomedical Informatics Research Network Data Repository . Neuroimage 124 : 1074 – 1079 . OpenUrl CrossRef PubMed 42. ↵ Du Y , Fu Z , Sui J , Gao S , Xing Y , Lin D , et al. ( 2020 ) : NeuroMark: An automated and adaptive ICA based pipeline to identify reproducible fMRI markers of brain disorders . Neuroimage Clin 28 : 102375 . OpenUrl CrossRef PubMed 43. ↵ Yao D , Sui J , Wang M , Yang E , Jiaerken Y , Luo N , et al. ( 2021 ): A Mutual Multi-Scale Triplet Graph Convolutional Network for Classification of Brain Disorders Using Functional or Structural Connectivity . IEEE Trans Med Imaging 40 : 1279 – 1289 . OpenUrl CrossRef PubMed 44. ↵ Li Y , Zhang X , Nie J , Zhang G , Fang R , Xu X , et al. ( 2022 ): Brain Connectivity Based Graph Convolutional Networks and Its Application to Infant Age Prediction . IEEE Trans Med Imaging 41 : 2764 – 2776 . OpenUrl PubMed 45. ↵ Hochreiter S , Schmidhuber J ( 1997 ): Long short-term memory . Neural Comput 9 . 46. ↵ Hochberg Y , Tamhane AC ( 1987 ): Multiple Comparison Procedures . John Wiley and Sons Inc . 47. ↵ Koutsouleris N , Davatzikos C , Borgwardt S , Gaser C , Bottlender R , Frodl T , et al. ( 2014 ) : Accelerated brain aging in schizophrenia and beyond: A neuroanatomical marker of psychiatric disorders . Schizophr Bull 40 : 1140 – 1153 . OpenUrl CrossRef PubMed 48. ↵ Fu Z , Sui J , Espinoza R , Narr K , Qi S , Sendi MSE , et al. ( 2022 ): Whole-Brain Functional Connectivity Dynamics Associated With Electroconvulsive Therapy Treatment Response . Biol Psychiatry Cogn Neurosci Neuroimaging 7 : 312 – 322 . OpenUrl PubMed 49. ↵ Schnack HG , Van Haren NEM , Nieuwenhuis M , Pol HEH , Cahn W , Kahn RS ( 2016 ) : Accelerated brain aging in schizophrenia: A longitudinal pattern recognition study . American Journal of Psychiatry 173 : 607 – 616 . OpenUrl CrossRef PubMed 50. ↵ Sendi MSE , Salat DH , Miller RL , Calhoun VD ( 2022 ): Two-step clustering-based pipeline for big dynamic functional network connectivity data . Front Neurosci 16 . 51. ↵ Shahab S , Mulsant BH , Levesque ML , Calarco N , Nazeri A , Wheeler AL , et al. ( 2019 ): Brain structure, cognition, and brain age in schizophrenia, bipolar disorder, and healthy controls . Neuropsychopharmacology 44 : 898 – 906 . OpenUrl CrossRef PubMed 52. ↵ Sendi MSE , Zendehrouh E , Sui J , Fu Z , Zhi D , Lv L , et al. ( 2021 ): Aberrant dynamic functional connectivity of default mode network predicts symptom severity in major depressive disorder . Brain Connect 11 : 838 – 849 . OpenUrl PubMed 53. ↵ Filley CM , Cullum CM ( 1994 ): Attention and vigilance functions in normal aging . Appl Neuropsychol 1 : 29 – 32 . OpenUrl PubMed 54. ↵ Nuechterlein KH , Green MF , Calkins ME , Greenwood TA , Gur RE , Gur RC , et al. ( 2015 ) : Attention/vigilance in schizophrenia: Performance results from a large multi-site study of the Consortium on the Genetics of Schizophrenia (COGS) . Schizophr Res 163 : 38 – 46 . OpenUrl CrossRef PubMed 55. ↵ Coull JT ( 2004 ): fMRI studies of temporal attention: Allocating attention within, or towards, time . Cognitive Brain Research 21 : 216 – 226 . OpenUrl CrossRef PubMed 56. ↵ Williams JHG , Waiter GD , Perra O , Perrett DI , Whiten A ( 2005 ): An fMRI study of joint attention experience . Neuroimage 25 : 133 – 140 . OpenUrl CrossRef PubMed Web of Science 57. ↵ Forbes NF , Carrick LA , McIntosh AM , Lawrie SM ( 2009 , June ) : Working memory in schizophrenia: A meta-analysis . Psychological Medicine , vol. 39 . pp 889 – 905 . OpenUrl CrossRef PubMed Web of Science 58. ↵ Wingfield A , Stine EAL , Lahar CJ , Aberdeen JS ( 1988 ): Does the capacity of working memory change with age? Exp Aging Res 14 : 103 – 107 . OpenUrl CrossRef PubMed Web of Science 59. ↵ Zhu J , Li Y , Fang Q , Shen Y , Qian Y , Cai H , Yu Y ( 2021 ): Dynamic functional connectome predicts individual working memory performance across diagnostic categories . Neuroimage Clin 30 . doi: 10.1016/j.nicl.2021.102593 OpenUrl CrossRef PubMed 60. ↵ Faghiri A , Thomas D , Oktay A , Calhoun VD ( 2020 ): A unified approach to study the brain connectivity frequency profile . International Symposium on Biomedical Imaging (ISBI) . 61. Hsu HC , Bai CH ( 2022 ): Individual and environmental factors associated with cognitive function in older people: a longitudinal multilevel analysis . BMC Geriatr 22 . doi: 10.1186/s12877-022-02940-9 OpenUrl CrossRef PubMed 62. ↵ Jeong HJ , Moore TM , Durham EL , Reimann GE , Dupont RM , Cardenas-Iniguez C , et al. ( 2023 ): General and Specific Factors of Environmental Stress and Their Associations With Brain Structure and Dimensions of Psychopathology . Biological Psychiatry Global Open Science 3 : 480 – 489 . OpenUrl PubMed 63. Puri S , Shaheen M , Grover B ( 2023 ): Nutrition and cognitive health: A life course approach . Frontiers in Public Health , vol. 11 . Frontiers Media S.A. doi: 10.3389/fpubh.2023.1023907 OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted July 22, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Functional Brain Age Acceleration from Dynamic and Static Connectivity Predicts Working Memory and Attention Deficits in Schizophrenia Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Functional Brain Age Acceleration from Dynamic and Static Connectivity Predicts Working Memory and Attention Deficits in Schizophrenia Sabrina J. Edwards-Swart , Bradley Baker , Daniel H. Mathalon , Judith M. Ford , Adrian Preda , Theo G.M. van Erp , Godfrey D. Pearlson , Jessica A. Turner , Vince D. Calhoun , Mohammad S. E. Sendi medRxiv 2025.07.21.25331956; doi: https://doi.org/10.1101/2025.07.21.25331956 Share This Article: Copy Citation Tools Functional Brain Age Acceleration from Dynamic and Static Connectivity Predicts Working Memory and Attention Deficits in Schizophrenia Sabrina J. Edwards-Swart , Bradley Baker , Daniel H. Mathalon , Judith M. Ford , Adrian Preda , Theo G.M. van Erp , Godfrey D. Pearlson , Jessica A. Turner , Vince D. Calhoun , Mohammad S. E. Sendi medRxiv 2025.07.21.25331956; doi: https://doi.org/10.1101/2025.07.21.25331956 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Psychiatry and Clinical Psychology Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4436) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15229) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6600) Geriatric Medicine (668) Health Economics (997) Health Informatics (4538) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (542) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3333) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9232) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a00eb375bfb741e2',t:'MTc3OTY1MDg4MA=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();