ReCOGnAIze app to detect mild cognitive impairment and vascular cognitive impairment

ReCOGnAIze app to detect mild cognitive impairment and vascular cognitive impairment | 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 ReCOGnAIze app to detect mild cognitive impairment and vascular cognitive impairment Adnan Azam Mohammed , Ashwati Vipin , Leow Yi Jin , Eliana Setiabudi , Farid Tan , Hitesh Agarwal , Kai Xin Liau , Pricilia Tanoto , Shan Yao Liew , Bocheng Qiu , Gurveen Kaur Sandhu , Jia Dong James Wang , Kiirtaara Aravindhan , Nagaendran Kandiah doi: https://doi.org/10.1101/2025.05.10.25327352 Adnan Azam Mohammed a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 b Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore Find this author on Google Scholar Find this author on PubMed Search for this author on this site Ashwati Vipin a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Leow Yi Jin a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Eliana Setiabudi a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Farid Tan a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Hitesh Agarwal a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kai Xin Liau a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Pricilia Tanoto a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Shan Yao Liew a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Bocheng Qiu a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Gurveen Kaur Sandhu a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jia Dong James Wang b Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kiirtaara Aravindhan a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 Find this author on Google Scholar Find this author on PubMed Search for this author on this site Nagaendran Kandiah a Dementia Research Centre (Singapore), Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 c Neuroscience and Mental Health Programme, Lee Kong Chian School of Medicine, Nanyang Technological University , Singapore 308232 d National Healthcare Group , 3 Fusionopolis Link, Nexus @ One-North, Singapore 138543 Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: Nagaendran_Kandiah{at}ntu.edu.sg Abstract Full Text Info/History Metrics Data/Code Preview PDF Abstract INTRODUCTION Mild Cognitive Impairment (MCI) is underdiagnosed due to subtle symptoms. Vascular Cognitive Impairment (VCI), the second leading cause of cognitive impairment, remains underdiagnosed due to varying non-amnestic manifestations. We aimed to develop and validate ReCOGnAIze, a gamified, interpretable digital app to detect MCI and VCI. METHODS A multi-phase, cross sectional study in an Asian community cohort with development phase (n=200) and validation with 235 independent participants having comprehensive neuroimaging and neuropsychological data. RESULTS In 235 subjects, ReCOGnAIze composite score accurately distinguished MCI from cognitively normal individuals (AUC = 0.90), outperforming the Montreal Cognitive Assessment (AUC = 0.70). For VCI detection, it achieved strong performance (n = 155; Average AUC = 0.85), identifying biomarkers of VCI; response time variability, task efficiency, consistent with known cerebrovascular disease impairments. DISCUSSION ReCOGnAIze is a scalable, explainable AI tool that accurately detects MCI and VCI, with interpretable clinical insights. 1. BACKGROUND Mild cognitive impairment (MCI) is widely recognized as a prodromal stage of dementia, with a significant 10–15% annual risk of conversion from MCI to dementia [ 1 , 2 ]. Given the global burden of dementia—currently affecting over 57 million individuals, projected to reach 150 million by 2050, the societal and economic impacts are profound [ 3 ]. With emerging pharmacological and non-pharmacological interventions demonstrating potential to delay dementia onset, even a modest delay of five years could reduce its prevalence by nearly 50%, underscoring the critical importance of early identification and intervention in MCI [ 4 , 5 ]. Vascular cognitive impairment (VCI), the second most common cause of cognitive decline globally, accounts for more than 20 million dementia cases [ 6 ]. Although there have been substantial advancements in pharmacotherapies for Alzheimer’s Disease (AD) with monoclonal antibodies targeting amyloid-β, these treatments may be less effective in VCI and carry increased risk of cerebral haemorrhage [ 7 ]. By contrast, VCI may respond to specific lifestyle interventions, and intensive management of vascular risk factors, such as hypertension, diabetes, and hyperlipidaemia [ 8 ]. The landmark SPRINT MIND study demonstrated that intensive blood pressure control could significantly reduce progression of cognitive impairment [ 9 ]. This highlights the importance of detecting and treating VCI accurately and early, to reduce the risk of future dementia and stroke. The burden of VCI is particularly high in Asian, African, and Hispanic populations, where elevated rates of hypertension, ischemic stroke, and diabetes drive cognitive decline [ 10 – 16 ]. A multi-country study across nine Asian cities highlighted the widespread prevalence of cerebrovascular disease and its detrimental impact on cognition [ 17 ]. Despite this, conventional cognitive screening tools such as the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) are optimized for memory deficits, limiting their utility in detecting hallmark VCI features, including slowed processing speed [ 18 ]. While comprehensive neuropsychological batteries for VCI exist, they are time-consuming and impractical for large-scale screening [ 19 , 20 ]. Digital cognitive assessments offer a promising solution to cognitive screening with self-administered, remote, repeatable, and automatically scored objective cognitive tests [ 21 ]. However, many are not specific for VCI. Commercially available platforms such as the CANTAB, Cognigram, and Neurotrack have primarily been validated for MCI of the AD type, rather than VCI [ 20 ]. Digital tools have rarely incorporated cerebrovascular markers in their development and validation, limiting their capability in differentiating between AD-related MCI and VCI [ 22 ]. Hence, there is an urgent need for screening tools to detect both MCI and VCI, given the rising burden of vascular risk factors, sedentary lifestyle and genetic predispositions making populations vulnerable to VCI [ 22 ]. To address this critical gap, we designed and developed the ReCOGnAIze App—a gamified digital cognitive assessment, for detecting MCI and VCI, validated in a community cohort having a high cerebrovascular burden. Gamified cognitive tasks with explainable AI models hold the potential to significantly enhance the early identification and differentiation of MCI and VCI, laying the groundwork for precise, scalable screening [ 8 , 9 ]. 2. METHODS ReCOGnAIze was designed, developed, and validated in two phases using data from participants in the Biomarker and Cognitive Study, Singapore (BIOCIS) [ 23 ]. 2.1 Study Population BIOCIS participants were classified as Cognitively Normal (CN) or as having MCI based on Peterson’s criteria and National Institute on Aging-Alzheimer’s Association guidelines after completion of a validated neuropsychological test battery [ 1 , 24 ]. Participants with MCI had a score of more than 4 on the Subjective Memory Questionnaire, had cognitive deficits 1.5 standard deviation (SD) below age matched norms, a global clinical dementia rating (CDR) score of 0.5-1, and had no functional deficits [ 25 ]. CN subjects had a global CDR of 0 and no significant neurocognitive deficits. Further methodological details are available in the BIOCIS protocol paper [ 23 , 26 ]. VCI was defined as per the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria and included cognitively impaired (either subjective or objective) individuals with neuroimaging evidence of vascular etiology identified by moderate-to-high levels of WMH on Brain-MRI (modified Fazekas score of 5 to 12). Non-vascular cognitively impaired (NVCI) included cognitively impaired (either subjective or objective) individuals with low levels of WMH on Brain-MRI (modified Fazekas score of 0 to 4) [ 6 , 11 , 27 ]. Two independent cohorts were used for development and validation. The development cohort comprised 200 subjects having demographic, neuropsychological and MRI data. The validation cohort included 235 subjects having demographic, neuropsychological and MRI data. All participants provided informed consent, and the study was approved by the institutional ethics board (IRB no: IRB-2021-1036). 2.2 Neuropsychological assessments and demographic data Briefly, demographic variables included age, sex, ethnicity, information about vascular risk factors: Diabetes Mellitus (DM), Hypertension (HTN), and Hyperlipidaemia (HLD) as defined by the clinical guidelines from the Ministry of Health, Singapore [ 28 – 30 ]. Neuropsychological measures included MoCA, Visual Cognitive Assessment Test (VCAT) [ 31 ], and composites Z scores of five cognitive domains: episodic memory (Rey Auditory Verbal Learning Test [RAVLT] delayed, Wechsler Memory Scale Fourth Edition [WMS-IV] Logical Story delayed, Rey-Osterrieth Complex Figure Test [RCFT] delayed), executive function (Colour Trails [CT] 2, Wechsler Adult Intelligence Scale Fourth Version [WAIS-IV] Digit span backward), processing speed (Trail Making Test [TMT] B, CT1, Symbol Digit Modalities Test [SDMT]), visuospatial (RCFT copy and WAIS-IV Block Design), and language domains (semantic fluency animals MoCA and semantic fluency vegetables VCAT) [ 32 – 39 ]. Participants underwent pen and paper neuropsychological testing facilitated by a trained researcher. Information about behaviour was collected using the mild behavioural impairment checklist (MBI-C) [ 40 ]. 2.3 Neuroimaging data Brain imaging was performed using the Siemens 3-Tesla Prisma MRI. The T1-weighted Magnetization Prepared Rapid Gradient Echo and Fluid Attenuated Inversion Recovery (FLAIR) sequences scan obtained was pre-processed using the Computational Anatomy Toolbox in the Statistical Parametric Mapping (SPM12) toolbox on MATLAB 2022b [ 41 ]. Measures of WMH through the Fazekas Rating and other measures of cerebral small vessel disease using the Staals’ criteria, were quantified and obtained by trained visual raters [ 42 , 43 ]. 2.4 Phase 1: Data-Guided Development of the App 2.4.1 Cognitive-behavioural profiles associated with MCI and VCI BIOCIS data from a sub-group of 200 subjects (Development Cohort) was analyzed [ 23 ] and multivariate analyses compared cognitive-behavioral profiles with MCI groups with moderate-to-high WMH and low WMH, controlling for age, education, APOE4 carrier status. MCI individuals with moderate-to-high WMH demonstrated significantly greater impairment in executive function, processing speed, and impulse control, with trending decreased motivation, measured by the mild behavioural impairment checklist (MBIC) [ 40 ], further shown in Figure 1 . These findings were important in developing a tool that can not only detect MCI, but also differentiate VCI. These were incorporated in cognitive game development. Download figure Open in new tab Figure 1. Cognitive-behavioral profile differences between MCI participants with and without VCI. Multivariate analysis of the development cohort (n = 200) comparing MCI participants with vascular cognitive impairment (VCI) to those without VCI (non-VCI) revealed significantly greater impairments in executive function, processing speed, and impulse control among the VCI group, after adjusting for age, education, and APOE4 carrier status. A trend toward reduced motivation was also observed, as measured by the Mild Behavioural Impairment Checklist (MBIC). These domain-specific deficits informed the selection and design of gamified cognitive tasks in the ReCOGnAIze app. 2.4.2 Development of gamified cognitive tasks Based on the cognitive-behavioural characterization, 4 gamified cognitive tasks were designed and developed as shown in Figure 2 . The tasks comprised of: Symbol Matching – based on the Digit Symbol Substitution Test intended to primarily assess Processing Speed; Trail Making – adapted version of the Trail Making Test intended to primarily assess Executive Function; Airplane Game – go/no-go paradigm intended to primarily assess Attention and Impulse Control; and Grocery Shopping – a Memory and Processing Speed assessment with increasing difficulty levels intended to assess memory, processing speed, impulse control and motivation [ 37 , 38 ]. The cognitive tasks are further described below: Download figure Open in new tab Figure 2. ReCOGnAIze Digital Cognitive Assessments and Their Targeted Domains. Four gamified, each designed to evaluate a distinct cognitive domain. Symbol Matching evaluates processing speed by requiring users to match abstract symbols to corresponding numbers. Trail Making evaluates executive function through an alternating attention task where users connect numbers within a green zone. Airplane Game evaluates attention and impulse control by asking users to swipe in the direction of a colored airplane while ignoring distractors. Grocery Shopping evaluates working memory and processing speed by memorizing a shopping list and selecting items from a conveyer belt in a limited time. 2.4.2.1 Symbol Matching (Processing Speed) A task requiring users to match symbols to their corresponding numbers within 1 minute with continuous randomization. This game traditionally measures information processing speed, which is commonly affected in patients with hallmark of subcortical ischemic VCI, where small vessel disease disrupts frontal-subcortical circuits. The game was chosen based on its strong literature backing as a sensitive marker of processing speed deficits [ 6 , 44 ]. Usability testing ensured that task complexity matched participants’ cognitive capabilities, refining the interface to optimize engagement and response accuracy. The game was designed with simple, intuitive controls and clear visual cues to ensure ease of interaction for users with varying levels of cognitive function. 2.4.2.2 Trail Making (Executive Function) An adaptation of the traditional trail-making test where users must connect all circles within 3 minutes, alternating between numbers and shapes, making it language-neutral. This task assesses executive function, cognitive flexibility, and set-shifting abilities, which are particularly vulnerable to frontal-subcortical circuit disruption in vascular cognitive impairment (VCI) [ 18 , 45 ]. Literature supports trail-making tests as a reliable indicator of executive dysfunction, making it a key inclusion in our app [ 35 , 37 , 46 , 47 ]. Focus groups provided feedback on time constraints, ensuring task difficulty was appropriate for older adults with varying cognitive abilities. The interface was designed to be highly responsive with clear step-by-step guidance, making it accessible even to individuals with mild motor impairments. 2.4.2.3 Airplane Game (Attention and Impulse Control) A go/no-go paradigm where users focus on the plane’s color and swipe in its pointing direction. This game measures response inhibition and impulse control [ 48 , 49 ]. Impulse dyscontrol is a significant behavioral manifestation in patients with high cerebrovascular burden. Selection of this game was further supported based on evidence supporting go/no-go paradigms in detecting inhibitory control deficits [ 50 , 51 ]. Usability testing refined task instructions and response timing to enhance accessibility and minimize errors unrelated to cognitive impairment. The game features a user-friendly design with large, easy-to-identify visual elements and minimal text requirements, making it suitable for users with varying literacy levels and cognitive abilities. 2.4.2.4 Grocery Shopping (Memory) A task requiring users to memorize a shopping list, select items from a conveyor belt, and prepare exact change. This game evaluates memory and attention. Selection of this task was supported by its ecological validity in assessing real-world memory function [ 52 , 53 ]. Focus groups provided feedback on task realism, improving user engagement and refining difficulty settings to balance challenge and feasibility. The game was designed with a visually engaging and interactive format to simulate a real-world shopping experience, ensuring it remains intuitive and enjoyable while accurately assessing memory function. 2.4.3 Integration of gamified cognitive tasks into a structured guided assessment app To ensure seamless and intuitive user experience, individual cognitive games were integrated into a structured framework with step-by-step instructions for each task, ensuring understanding and completion without external assistance. Each game was accompanied by guided tutorials, visual cues, text-based prompts, and interactive practice trials to familiarize users with the mechanics before starting the assessment. The assessment followed a Learning -> Trial -> Assess framework. Users first engage in a guided learning phase where they are introduced to the task through step-by-step explanations and demonstrations. This is followed by a trial phase, allowing users to practice in a low-stakes environment with real-time feedback to build confidence and familiarity. Finally, the test phase objectively measures cognitive performance under standardized conditions, ensuring valid and reliable results. User feedback from pilot testing and focus groups informed iterative refinements in the instruction delivery and task order. Adjustments were made to optimize user comprehension, minimize errors due to misunderstanding, and enhance the overall experience of taking the assessment. Through these enhancements, the ReCOGnAIze app was transformed into a comprehensive, structured cognitive assessment tool that is both scientifically robust and highly accessible. 2.5 Phase 2: Scoring and Validation of the App 2.5.1 Study Population (Validation Cohort) An independent cohort of 235 participants was recruited from the BIOCIS study for testing and validation of ReCOGnAIze for detection of MCI and differentiation of VCI from NVCI [ 23 ]. Participants underwent the assessment on the ReCOGnAIze web application. Data collected was further processed as detailed below. 2.5.2 Composite Cognitive Performance Scores To validate ReCOGnAIze’s cognitive assessment ability, we calculated a normalized, equally-weighted composite score Recognaize _ total to reflect global cognition. Each of the four games in ReCOGnAIze was used to generate an individual score based on accuracy of task completion and response time, computed as: These were normalized using Min-Max scaling, mapping each score to a range between 0 and 5: where Normalized_Game_Score 1 is the normalized score for game i , and min( Game_Score 1 )and max( Game_Score 1 ) represent the minimum and maximum raw scores observed in the sample for that game. The ReCOGnAIze Composite Score was then calculated as the summation of the four normalized game scores: 2.5.3 Feature Engineering and Selection for VCI Feature engineering was done to extract granular game features of attention, processing speed and executive function: time-based, accuracy, performance consistency, and efficiency, to enable a deeper investigation into the pathobiology of VCI. Time-based measures capture response speed and variability, reflecting cognitive efficiency and flexibility. Central tendency statistics (mean, median, mode) determined typical response speed, while range measures (minimum, maximum) highlighted performance fluctuations. Variability metrics (standard deviation, interquartile range (IQR), CV) assessed response stability, as greater variability could indicate cognitive decline. Percentile measures (25th, 75th) indicated shifts in response distribution, and time trend analyses examined changes over the task, revealing fatigue, learning effects, or attentional lapses. Accuracy measures were used to assess task performance and error patterns. Accuracy changes between the first and second halves of tasks could identify cognitive fatigue or adaptation. Performance consistency reflected cognitive stability and resilience, with sustained attention measures like longest streak of correct or incorrect responses. Efficiency measures integrated performance indices and efficiency scores to identify compensation strategies. To determine the most predictive, interpretable digital cognitive features for VCI, we applied a three-stage pipeline. First, we computed SHapley Additive Explanations (SHAP) values from a Random Forest model and retained the top 30 most important features [ 54 ]. Second, we applied Elastic Net regularization to select robust predictors while accounting for multicollinearity between features [ 55 ]. Third, we performed Recursive Feature Elimination with cross-validation (RFECV) using logistic regression to identify the minimal subset of features with maximal predictive performance for VCI [ 56 ]. 2.5.4 Vascular Cognitive Scores Selected features were combined with demographic and vascular risk factors associated with VCI — Age, Diabetes Mellitus (DM), Hypertension (HTN), and Hyperlipidaemia (HLD) [ 57 ]—to construct Vascular Cognitive Scores (VCS) using both rule-based and ML approaches. We computed Recognaize_VCS Composite , a normalized, equally-weighted mean of selected cognitive features and risk variables: where F n represents the selected digital cognitive features. ML classifiers: Random Forest, XGBoost, LightGBM, and CatBoost were also trained to detect VCI, yielding ML-derived VCS which were validated with 5 fold cross validation (CV) in detecting VCI. Mathematically: where M {- j } denotes the ML model trained on all data excluding fold j , and superscript ( j ) indicates the data of the held-out fold. For unbiased and generalizable models, all ML-based VCS were generated using out-of-fold (OOF) predictions from a 5-fold Cross-Validation (CV) procedure. Each model was trained on 80% of the dataset and predict VCI for the held-out 20%. This process was repeated across all five folds such that each subject’s VCS was obtained from a model that had not been trained on their data. This approach ensures that no subject’s score was influenced by a model trained on their own data, reducing overfitting risk and improving downstream analyses’ robustness. 2.6 Statistical Methods All statistical analyses were conducted using Python (v3.9, Scipy, Scikit-learn, CatBoost, LightGBM, XGBoost, SHAP, ElasticNet, RFECV) [ 54 – 56 , 58 – 63 ]. 2.6.1 Participant Characteristics Baseline characteristics of the study cohort were summarized using means and SD for continuous variables. Categorical variables were presented as percentages. Group comparisons between CN and MCI individuals, and between VCI and NVCI individuals, were performed using independent t-tests for continuous variables and chi-squared tests for categorical variables. For cognitive outcome measures, additional age- and education-adjusted comparisons were performed using analysis of covariance (ANCOVA), with diagnostic group as the between-subjects factor and age and years of education as covariates. Adjusted means ± standard errors (SE), p values, and partial eta squared (eta²) effect sizes were reported. Cohen’s d was calculated for unadjusted comparisons to estimate effect sizes. All p values were two tailed, and a threshold of p <0.05 was considered statistically significant. 2.6.2 Differentiating MCI from CN The ability of the Recognaize_total composite score to differentiate MCI from CN was assessed using receiver operator characteristic (ROC) curves, with area under the curve (AUC), 95% confidence interval, sensitivity, specificity reported. To benchmark diagnostic performance, we compared the Recognaize_total Composite Score against established MCI screening tools: MoCA and VCAT [ 31 , 64 ]. The optimal cut-off for MCI detection was determined using Youden’s J statistic [ 65 ]. 2.6.3 Differentiating VCI from NVCI The ability of the different VCS scores to differentiate VCI from NVCI was assessed using 5-fold CV average AUC with accuracy, sensitivity, specificity and F1-score (for ML models) reported. 3. RESULTS 3.1 Participant Characteristics As shown in Table 1 , from 235 total, MCI participants (n = 108; mean age, 65.0 years [SD 9.0]) were significantly older than CN (n = 127; 58.2 years [SD 9.6]; p < .001, d = 0.74). MCI participants also had fewer years of education (13.9 [SD 3.4] vs 15.1 [SD 2.8]; p = .004, d = 0.39). In age- and education-adjusted analyses, MCI participants had significantly lower MoCA scores (adjusted mean, 25.16 [SE 0.24] vs 26.59 [SE 0.22]; p < .001; partial eta² = 0.07) and VCAT scores (26.31 [SE 0.24] vs 27.20 [SE 0.22]; p = .009; partial eta² = 0.03), and a large group difference in executive function (–0.67 [SE 0.06] vs –0.00 [SE 0.06]; p < .001; partial eta² = 0.19) ( Table 1 ).The ReCOGnAIze_total composite showed the largest group difference (adjusted mean, 7.72 [SE 0.17] vs 10.47 [SE 0.16] ; p < .001; partial eta² = 0.36), outperforming MoCA (eta² = 0.07) and VCAT (eta ² = 0.03). ReCOGnAIze’s individual game scores also significantly differed between groups with large effect sizes: Trail Making (eta² = 0.18), Airplane Game (eta² = 0.16), Grocery Shopping (eta² = 0.18), and Symbol Matching (eta² = 0.13) ( Table 1 ). View this table: View inline View popup Table 1. Comparison of Demographic Characteristics and Cognitive Scores Between Cognitively Normal and Mild Cognitive Impairment Groups 3.2 Detection of Mild Cognitive Impairment As shown in Figure 3 , ROC analyses comparing MCI and CN demonstrated that Recognaize_total composite score achieved the highest diagnostic accuracy (AUC = 0.90, 95% CI 0.87–0.94) with a sensitivity of 0.85 and specificity of 0.84 at the optimal threshold detailed in Table 2 . This outperformed traditional tools, MoCA (AUC = 0.70, 95% CI 0.63– 0.77) and VCAT (AUC = 0.66, 95% CI 0.59–0.73) as shown in Figure 3 , Panel A . Download figure Open in new tab Figure 3. Diagnostic accuracy of ReCOGnAIze composite score, and individual game scores, compared to traditional tests, in detecting MCI. Panel A shows ROC curves for ReCOGnAIze total score, Montreal Cognitive Assessment (MoCA), and Visual Cognitive Assessment Test (VCAT) in distinguishing individuals with mild cognitive impairment (MCI) from cognitively normal (CN) individuals. Panel B displays ROC curves for the ReCOGnAIze total score and individual game scores: Trail Making, Grocery Shopping, Symbol Matching, and Airplane Game. Circles indicate optimal operating points based on the Youden index. Area under the curve (AUC) values with 95% confidence intervals (CIs) are reported in the legend. ReCOGnAIze outperformed traditional tools and demonstrated strong discriminative ability across both total and individual game scores. View this table: View inline View popup Download powerpoint Table 2. Diagnostic performance of ReCOGnAIze, its subtests, the Montreal Cognitive Assessment and the Visual Cognitive Assessment Test for detecting Mild Cognitive Impairment with area under the curve, sensitivity, specificity, and optimal threshold reported Figure 3 , Panel B shows that individual ReCOGnAIze games also demonstrated robust classification performance: Trail Making (AUC = 0.81, 95% CI 0.75–0.86), Grocery Shopping (AUC = 0.79, 95% CI 0.74–0.85), Symbol Matching (AUC = 0.78, 95% CI 0.73–0.84), and Airplane Game (AUC = 0.78, 95% CI 0.71–0.83). AUCs, optimal thresholds, specificity and sensitivity for all tests are provided in Table 2 . These findings support the validity of ReCOGnAIze and its component tasks in detecting MCI. 3.3 Detection of Vascular Cognitive Impairment As shown in Table 3 , participants with VCI (n = 75; mean age, 67.1 years [SD 7.6]) were significantly older than those without VCI (n = 79; 60.6 years [SD 9.0]; p < .001, d = 0.78). Hyperlipidaemia was significantly more common in the VCI group. White matter disease burden was higher in the VCI group, with higher Fazekas Total scores and WMH volumes. Machine learning–based Vascular Cognitive Scores showed the strongest separation between groups even after adjusting for age: CatBoost-VCS (adjusted mean, 0.62 [SE 0.02] vs 0.39 [SE 0.02]; p < .001; eta² = 0.22) and XGBoost-VCS (0.66 [SE 0.04] vs 0.29 [SE 0.04]; p < .001; eta² = 0.22), outperforming MoCA and VCAT. View this table: View inline View popup Table 3. Comparison of Demographic Characteristics, Imaging Markers, Cognitive Scores and ML-Derived Vascular Cognitive Scores between VCI and NVCI sub-groups The highest average AUC for detecting VCI was observed with the CatBoost (AUC 0.85) and XGBoost (AUC 0.84) models, detailed in Table 4 . Both models outperformed MoCA (AUC 0.65) and the Recognaize-VCS Composite score (AUC 0.77), with CatBoost also achieving the highest accuracy (0.85), sensitivity (0.85), and F1 score (0.82). View this table: View inline View popup Download powerpoint Table 4. Diagnostic performance of machine learning–based Vascular Cognitive Scores (VCS) and standard tools for detecting VCI. Novel digital cognitive features predictive of VCI Feature importance ranking using SHAP values for the CatBoost model identified the most predictive features of VCI as gs1_avg_success_time [Grocery Shopping Average Success Time] (mean SHAP value +1.15) and sm_time_iqr [Symbol Matching IQR](+0.88), followed by age (+0.64), variability in round time, and total task completion time, as shown in Figure 4 . These findings indicate that fine-grained, time-based behavioral metrics from cognitive game performance provided stronger discriminatory power than traditional cognitive test scores, especially in cerebrovascular cognitive impairments. Download figure Open in new tab Figure 4. Feature importance ranking based on SHapley Additive exPlanations (SHAP) values for the CatBoost classification model used to distinguish VCI from NVCI. Higher mean absolute SHAP values indicate greater contribution to the model’s predictions. Key predictive features revealed novel digital biomarkers of VCI such as Grocery Shopping average success time (gs1_avg_success_time) and Symbol Matching response time variability (sm_time_iqr), which aligns with known clinical findings of executive dysfunction and reduced processing speed in cerebrovascular disease. 4. DISCUSSION In this study, we developed and validated ReCOGnAIze, a gamified, AI-powered digital cognitive assessment tool designed to detect both MCI and VCI. Grounded in cognitive-behavioral science, the tool incorporates engaging tasks with automated scoring and interpretable ML models. We validated a Recognaize_total composite score which achieved an AUC of 0.9 for distinguishing MCI from CN, outperforming MoCA and VCAT, each of which achieved AUC less than 0.7. In addition to detecting MCI, we used explainable AI for identifying digital cognitive features for differentiating VCI from NVCI, revealing key insights about cerebrovascular-driven cognitive impairment. We then validated ML models for detecting VCI from NVCI, using ReCOGnAIze features and VCI risk variables, achieving an average AUC of 0.85. These findings suggest that ReCOGnAIze may serve as an effective and scalable cognitive screening tool, with interpretability and clinical relevance for identifying both MCI and VCI. Compared to established tools such as MoCA, which reports AUCs ranging from 0.70 to 0.85 across diverse populations[ 64 ], ReCOGnAIze achieved comparable or higher AUCs while targeting cognitive domains underrepresented in traditional assessments. These include processing speed, response inhibition, and sustained attention—domains commonly impaired in VCI and non-amnestic MCI subtypes [ 18 ]. This task-specific evaluation might explain its high discriminative accuracy. The app’s brief duration (10–15 minutes), gamified design, and automated output enhance its potential for large-scale screening. While other digital tools have shown promise in detecting MCI, few have been validated specifically against neuroimaging or cerebrovascular biomarkers which show associations with non-amnestic domain impairments. A key strength of ReCOGnAIze lies in its capability to detect VCI—an area largely overlooked by existing digital platforms despite the high global prevalence. Our explainable AI driven feature selection process highlighted novel digital biomarkers: response time variability, impulse control, and task switching—in line with executive impairments associated with VCI [ 6 , 18 ]. To the best of our knowledge, this is one of the first digital tools for the targeted detection of vascular-specific cognitive impairment using explainable feature attribution. With explainable AI and clinical domain expertise, we prioritized model transparency and interpretability without compromising performance. After cross-validation, we used SHAP to generate individualized explanations of the predictions, facilitating clinician understanding and trust. While several studies have applied black-box models to cognitive classification, few have incorporated interpretable AI in a clinically meaningful way. Our pipeline ensures that each retained feature—such as response time variability, inhibition errors—has both a statistical and neuropsychological rationale, strengthening its relevance for clinical deployment in screening for MCI and VCI. Key strengths of this study include the use of a deeply phenotyped, community based cohort with Brain-MRI, integration of cognitive-behavioural science in game design, and application of rigorous, multi-stage ML. Cognitive validity was established using a generalizable composite score, and ML models for differential detection were evaluated using cross-validation and out-of-fold predictions, reducing overfitting risk. Nonetheless, several limitations must be acknowledged. First, the cross-sectional nature of the study precludes evaluation of predictive validity for future cognitive decline or dementia conversion. Longitudinal studies are needed to determine prognostic utility. Second, external validation in diverse, low-literacy, and rural populations is required. Third, cerebrovascular burden was assessed using WMH severity, which—while clinically relevant—does not capture all aspects of cerebrovascular pathology. Finally, performance in real-world clinical or community settings remains to be evaluated. The findings of this study have several implications for clinical practice and public health. ReCOGnAIze offers a scalable, self-administered digital cognitive assessment that could be deployed in primary care or community settings for cognitive screening. Its short duration, engaging interface, and minimal training requirements make it feasible for wide-scale use. Importantly, the tool evaluates VCI—an under-recognized but potentially treatable contributor to dementia, with a high global burden. Conclusion ReCOGnAIze is a novel, interpretable digital cognitive assessment that enables accurate detection of both MCI and VCI. By combining gamified tasks with explainable AI, it offers a transparent, efficient, and clinically relevant approach to early cognitive screening. This tool holds promise for transforming dementia risk stratification in both clinical and population health contexts, advancing precision medicine approaches to cognitive care. Data Availability Selective data produced in the present study can be made available upon reasonable request to the authors and corresponding author. AUTHOR CONTRIBUTIONS AAM contributed to the conception, study design, data acquisition and analysis and drafting of the manuscript and figures. AV contributed to the study design, data acquisition and drafting of the manuscript. YJL contributed to the data acquisition and drafting of the manuscript. ES contributed to data acquisition and drafting of the manuscript and figures. FT contributed to data acquisition and drafting of the manuscript. HA contributed to data acquisition and drafting of the manuscript. KXL contributed to data acquisition and drafting of the manuscript. PT contributed to data acquisition and drafting of the manuscript. GKS contributed to data acquisition and drafting of the manuscript. JDJW contributed to data acquisition and drafting of the manuscript. BQ contributed to data acquisition and drafting of the manuscript. KA contributed to data acquisition and drafting of the manuscript. NK contributed to conception, study design, data acquisition and analysis and drafting of the manuscript. CONFLICTS OF INTEREST All authors declare they have no competing interests. DATA SHARING STATEMENT Data used in this study will be shared upon reasonable request made to the Corresponding Author. ACKNOWLEDGEMENTS, SOURCES OF FUNDING AND DISCLOSURES FUNDING SOURCES This study received funding support from the Strategic Academic Initiative grant (SP1CLNT900-NTU-A630-PJ-03INP001400A630) from the Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore, National Medical Research Council, Singapore under its Clinician Scientist Award (MOH-CSAINV18nov-0007), Ministry of Education Start-up Grant, Ministry of Education Academic Research Fund Tier 1 (RT02/21) and Ministry of Education Science of Learning grant (MOESOL2022-0002), NTUitive GAP Fund from Nanyang Technological University, Singapore (NGF-2023-13-019) and the Innovation to Startup Grant (I2Start-2307024) from the Singapore-MIT Alliance for Research and Technology (SMART) under the National Research Foundation (NRF), Singapore. REFERENCES [1]. ↵ Petersen RC , Knopman DS , Boeve BF , Geda YE , Ivnik RJ , Smith GE , et al. Mild Cognitive Impairment: Ten Years Later . Arch Neurol 2009 ; 66 : 1447 – 55 . doi: 10.1001/archneurol.2009.266 . OpenUrl CrossRef PubMed Web of Science [2]. ↵ Petersen RC , Lopez O , Armstrong MJ , Getchius TSD , Ganguli M , Gloss D , et al. Practice guideline update summary: Mild cognitive impairment: Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology . Neurology 2018 ; 90 : 126 – 35 . doi: 10.1212/WNL.0000000000004826 . OpenUrl CrossRef PubMed [3]. ↵ Wimo A , Seeher K , Cataldi R , Cyhlarova E , Dielemann JL , Frisell O , et al. The worldwide costs of dementia in 2019 . Alzheimers Dement 2023 ; 19 : 2865 – 73 . doi: 10.1002/alz.12901 . OpenUrl CrossRef [4]. ↵ Zissimopoulos JM , Tysinger BC , St.Clair PA , Crimmins EM. The Impact of Changes in Population Health and Mortality on Future Prevalence of Alzheimer’s Disease and Other Dementias in the United States . J Gerontol B Psychol Sci Soc Sci 2018 ; 73 : S38 – 47 . doi: 10.1093/geronb/gbx147 . OpenUrl CrossRef PubMed [5]. ↵ Michel J-P . Is It Possible to Delay or Prevent Age-Related Cognitive Decline? Korean J Fam Med 2016 ; 37 : 263 – 6 . doi: 10.4082/kjfm.2016.37.5.263 . OpenUrl CrossRef PubMed [6]. ↵ Gorelick PB , Scuteri A , Black SE , Decarli C , Greenberg SM , Iadecola C , et al. Vascular contributions to cognitive impairment and dementia: a statement for healthcare professionals from the american heart association/american stroke association . Stroke 2011 ; 42 : 2672 – 713 . doi: 10.1161/STR.0b013e3182299496 . OpenUrl Abstract / FREE Full Text [7]. ↵ Dyck CH van , Swanson CJ , Aisen P , Bateman RJ , Chen C , Gee M , et al. Lecanemab in Early Alzheimer’s Disease . New England Journal of Medicine 2023 ; 388 : 9 – 21 . doi: 10.1056/NEJMoa2212948 . OpenUrl CrossRef PubMed [8]. ↵ The SPRINT MIND Investigators for the SPRINT Research Group. Effect of Intensive vs Standard Blood Pressure Control on Probable Dementia: A Randomized Clinical Trial . JAMA 2019 ; 321 : 553 – 61 . doi: 10.1001/jama.2018.21442 . OpenUrl CrossRef PubMed [9]. ↵ Elahi FM , Alladi S , Black SE , Claassen JAHR , DeCarli C , Hughes TM , et al. Clinical trials in vascular cognitive impairment following SPRINT-MIND: An international perspective . Cell Rep Med 2023 ; 4 : 101089 . doi: 10.1016/j.xcrm.2023.101089 . OpenUrl CrossRef PubMed [10]. ↵ Chen JXY , Vipin A , Sandhu GK , Leow YJ , Zailan FZ , Tanoto P , et al. Blood-brain barrier integrity disruption is associated with both chronic vascular risk factors and white matter hyperintensities . The Journal of Prevention of Alzheimer’s Disease 2025 ; 12 : 100029 . doi: 10.1016/j.tjpad.2024.100029 . OpenUrl CrossRef [11]. ↵ Vipin A , Satish V , Saffari SE , Koh W , Lim L , Silva E , et al. Dementia in Southeast Asia: influence of onset-type, education, and cerebrovascular disease . Alzheimer’s Research & Therapy 2021 ; 13 : 195 . doi: 10.1186/s13195-021-00936-y . OpenUrl CrossRef [12]. Leow YJ , Soo SA , Kumar D , Zailan FZB , Sandhu GK , Vipin A , et al. Mild Behavioral Impairment and Cerebrovascular Profiles Are Associated with Early Cognitive Impairment in a Community-Based Southeast Asian Cohort . J Alzheimers Dis 2024 ; 97 : 1727 – 35 . doi: 10.3233/JAD-230898 . OpenUrl CrossRef PubMed [13]. Bhalla G , Tanoto P , Vipin A , Chen XYJ , Leow YJ , Chen C , et al. Current status and future directions for the diagnosis and management of mild cognitive impairment in Southeast Asia: A SEACURE consensus paper . J Prev Alzheimers Dis 2025 : 100110 . doi: 10.1016/j.tjpad.2025.100110 . OpenUrl CrossRef [14]. Schneider BC , Gross AL , Bangen KJ , Skinner JC , Benitez A , Glymour MM , et al. Association of Vascular Risk Factors With Cognition in a Multiethnic Sample . J Gerontol B Psychol Sci Soc Sci 2015 ; 70 : 532 – 44 . doi: 10.1093/geronb/gbu040 . OpenUrl CrossRef PubMed [15]. Williams IC , Park MH , Tsang S , Sperling SA , Manning C . Cognitive Function and Vascular Risk Factors Among Older African American Adults . J Immigr Minor Health 2018 ; 20 : 612 – 8 . doi: 10.1007/s10903-017-0583-7 . OpenUrl CrossRef PubMed [16]. ↵ Becker CJ , Heeringa SG , Chang W , Briceño EM , Mehdipanah R , Levine DA , et al. Differential Impact of Stroke on Cognitive Impairment in Mexican Americans and Non-Hispanic White Americans . Stroke 2022 ; 53 : 3394 – 400 . doi: 10.1161/STROKEAHA.122.039533 . OpenUrl CrossRef PubMed [17]. ↵ High burden of cerebral white matter lesion in 9 Asian cities | Scientific Reports n.d. https://www.nature.com/articles/s41598-021-90746-x (accessed March 23, 2025). [18]. ↵ O’Brien JT , Thomas A. Vascular dementia . The Lancet 2015 ; 386 : 1698 – 706 . doi: 10.1016/S0140-6736(15)00463-8 . OpenUrl CrossRef PubMed [19]. ↵ Ferris SH . General measures of cognition . Int Psychogeriatr 2003 ; 15 Suppl 1 : 215 – 7 . doi: 10.1017/S1041610203009220 . OpenUrl CrossRef PubMed [20]. ↵ Khaw J , Subramaniam P , Abd Aziz NA , Ali Raymond A , Wan Zaidi WA , Ghazali SE . Current Update on the Clinical Utility of MMSE and MoCA for Stroke Patients in Asia: A Systematic Review . Int J Environ Res Public Health 2021 ; 18 : 8962 . doi: 10.3390/ijerph18178962 . OpenUrl CrossRef PubMed [21]. ↵ Chan JYC , Yau STY , Kwok TCY , Tsoi KKF . Diagnostic performance of digital cognitive tests for the identification of MCI and dementia: A systematic review . Ageing Research Reviews 2021 ; 72 : 101506 . doi: 10.1016/j.arr.2021.101506 . OpenUrl CrossRef PubMed [22]. ↵ Salvadori E , Pantoni L . Teleneuropsychology for vascular cognitive impairment: Which tools do we have? Cereb Circ Cogn Behav 2023 ; 5 : 100173 . doi: 10.1016/j.cccb.2023.100173 . OpenUrl CrossRef PubMed [23]. ↵ Leow YJ , Wang JDJ , Vipin A , Sandhu GK , Soo SA , Kumar D , et al. Biomarkers and Cognition Study, Singapore (BIOCIS): Protocol, Study Design, and Preliminary Findings . J Prev Alzheimers Dis 2024 ; 11 : 1093 – 105 . doi: 10.14283/jpad.2024.89 . OpenUrl CrossRef PubMed [24]. ↵ Albert MS , DeKosky ST , Dickson D , Dubois B , Feldman HH , Fox NC , et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease . Alzheimers Dement 2011 ; 7 : 270 – 9 . doi: 10.1016/j.jalz.2011.03.008 . OpenUrl CrossRef PubMed Web of Science [25]. ↵ Bennett-Levy J , Powell GE . The Subjective Memory Questionnaire (SMQ). An investigation into the self-reporting of ‘real-life’ memory skills . British Journal of Social and Clinical Psychology 1980 ; 19 : 177 – 88 . doi: 10.1111/j.2044-8260.1980.tb00946.x . OpenUrl CrossRef [26]. ↵ Morris JC . Clinical dementia rating: a reliable and valid diagnostic and staging measure for dementia of the Alzheimer type . Int Psychogeriatr 1997 ; 9 Suppl 1 : 173 – 6 ; discussion 177-178. doi: 10.1017/s1041610297004870 . OpenUrl CrossRef PubMed [27]. ↵ Diagnostic and Statistical Manual of Mental Disorders | Psychiatry Online. DSM Library n.d. https://psychiatryonline.org/doi/book/10.1176/appi.books.9780890425787 (accessed April 15, 2025). [28]. ↵ Tai E , Chia B , Bastian A , Chua T , Ho S , Koh T , et al. Ministry of Health Clinical Practice Guidelines: Lipids . Smedj 2017 ; 58 : 155 – 66 . doi: 10.11622/smedj.2017018 . OpenUrl CrossRef [29]. Tay J , Sule A , Chew E , Tey J , Lau T , Lee S , et al. Ministry of Health Clinical Practice Guidelines: Hypertension . Smedj 2018 ; 59 : 17 – 27 . doi: 10.11622/smedj.2018007 . OpenUrl CrossRef [30]. ↵ MOH | Guidelines n.d. https://www.hpp.moh.gov.sg/doctors/guidelines/guidelinedetails/cpgmed_diabetes_mellitus (accessed April 21, 2025). [31]. ↵ Soo SA , Kumar D , Leow YJ , Koh CL , Saffari SE , Kandiah N . Usefulness of the Visual Cognitive Assessment Test in Detecting Mild Cognitive Impairment in the Community . Journal of Alzheimer’s Disease 2023 ; 93 : 755 – 63 . doi: 10.3233/JAD-221301 . OpenUrl CrossRef [32]. ↵ Brief Informant Screening Test for Mild Cognitive Impairment and Early Alzheimer’s Disease | Dementia and Geriatric Cognitive Disorders | Karger Publishers n.d. https://karger.com/dem/article-abstract/21/5-6/392/97942/Brief-Informant-Screening-Test-for-Mild-Cognitive (accessed April 16, 2025). [33]. Wechsler D . Wechsler memory scale . San Antonio, TX, US : Psychological Corporation ; 1945 . [34]. Zhang X , Lv L , Min G , Wang Q , Zhao Y , Li Y . Overview of the Complex Figure Test and Its Clinical Application in Neuropsychiatric Disorders, Including Copying and Recall . Front Neurol 2021 ; 12 : 680474 . doi: 10.3389/fneur.2021.680474 . OpenUrl CrossRef PubMed [35]. ↵ Tyburski E , Karabanowicz E , Mak M , Lebiecka Z , Samochowiec A , Pełka-Wysiecka J , et al. Color Trails Test: A New Set of Data on Cognitive Flexibility and Processing Speed in Schizophrenia . Front Psychiatry 2020 ; 11 : 521 . doi: 10.3389/fpsyt.2020.00521 . OpenUrl CrossRef PubMed [36]. Ruchinskas R . Wechsler adult intelligence scale-4th edition digit span performance in subjective cognitive complaints, amnestic mild cognitive impairment, and probable dementia of the Alzheimer type . Clin Neuropsychol 2019 ; 33 : 1436 – 44 . doi: 10.1080/13854046.2019.1585574 . OpenUrl CrossRef PubMed [37]. ↵ Kortte KB , Horner MD , Windham WK . The trail making test, part B: cognitive flexibility or ability to maintain set? Appl Neuropsychol 2002 ; 9 : 106 – 9 . doi: 10.1207/S15324826AN0902_5 . OpenUrl CrossRef PubMed [38]. ↵ Charvet LE , Beekman R , Amadiume N , Belman AL , Krupp LB . The Symbol Digit Modalities Test is an effective cognitive screen in pediatric onset multiple sclerosis (MS) . J Neurol Sci 2014 ; 341 : 79 – 84 . doi: 10.1016/j.jns.2014.04.006 . OpenUrl CrossRef PubMed [39]. ↵ Reverberi C , Cherubini P , Baldinelli S , Luzzi S . Semantic fluency: cognitive basis and diagnostic performance in focal dementias and Alzheimer’s disease . Cortex 2014 ; 54 : 150 – 64 . doi: 10.1016/j.cortex.2014.02.006 . OpenUrl CrossRef PubMed [40]. ↵ Ismail Z , Agüera-Ortiz L , Brodaty H , Cieslak A , Cummings J , Fischer CE , et al. The Mild Behavioral Impairment Checklist (MBI-C): A Rating Scale for Neuropsychiatric Symptoms in Pre-Dementia Populations . J Alzheimers Dis 2017 ; 56 : 929 – 38 . doi: 10.3233/JAD-160979 . OpenUrl CrossRef PubMed [41]. ↵ Gaser C , Dahnke R , Thompson PM , Kurth F , Luders E , The Alzheimer’s Disease Neuroimaging Initiative null. CAT: a computational anatomy toolbox for the analysis of structural MRI data . Gigascience 2024 ; 13 : giae049 . doi: 10.1093/gigascience/giae049 . OpenUrl CrossRef PubMed [42]. ↵ Staals J , Booth T , Morris Z , Bastin ME , Gow AJ , Corley J , et al. Total MRI load of cerebral small vessel disease and cognitive ability in older people . Neurobiol Aging 2015 ; 36 : 2806 – 11 . doi: 10.1016/j.neurobiolaging.2015.06.024 . OpenUrl CrossRef PubMed [43]. ↵ Wardlaw JM , Smith EE , Biessels GJ , Cordonnier C , Fazekas F , Frayne R , et al. Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration . Lancet Neurol 2013 ; 12 : 822 – 38 . doi: 10.1016/S1474-4422(13)70124-8 . OpenUrl CrossRef PubMed Web of Science [44]. ↵ Jaeger J . Digit Symbol Substitution Test . J Clin Psychopharmacol 2018 ; 38 : 513 – 9 . doi: 10.1097/JCP.0000000000000941 . OpenUrl CrossRef PubMed [45]. ↵ Dichgans M , Leys D. Vascular Cognitive Impairment . Circulation Research 2017 ; 120 : 573 – 91 . doi: 10.1161/CIRCRESAHA.116.308426 . OpenUrl Abstract / FREE Full Text [46]. ↵ Trail Making Test - an overview | ScienceDirect Topics n.d. https://www.sciencedirect.com/topics/medicine-and-dentistry/trail-making-test (accessed April 21, 2025). [47]. ↵ Linari I , Juantorena GE , Ibáñez A , Petroni A , Kamienkowski JE . Unveiling Trail Making Test: visual and manual trajectories indexing multiple executive processes . Sci Rep 2022 ; 12 : 14265 . doi: 10.1038/s41598-022-16431-9 . OpenUrl CrossRef PubMed [48]. ↵ Kreutzer JS , DeLuca J , Caplan B Iverson GL . Go/No-Go Testing . In: Kreutzer JS , DeLuca J , Caplan B , editors. Encyclopedia of Clinical Neuropsychology , New York, NY : Springer; 2011 , p. 1162 – 3 . doi: 10.1007/978-0-387-79948-3_185 . OpenUrl CrossRef [49]. ↵ Meule A . Reporting and Interpreting Task Performance in Go/No-Go Affective Shifting Tasks . Front Psychol 2017 ; 8 . doi: 10.3389/fpsyg.2017.00701 . OpenUrl CrossRef [50]. ↵ Kim S , Lee D. Prefrontal Cortex and Impulsive Decision Making . Biol Psychiatry 2011 ; 69 : 1140 – 6 . doi: 10.1016/j.biopsych.2010.07.005 . OpenUrl CrossRef PubMed Web of Science [51]. ↵ Poststroke Impulsivity: A Narrative Review | The Journal of Neuropsychiatry and Clinical Neurosciences n.d. https://psychiatryonline.org/doi/10.1176/appi.neuropsych.20240080 (accessed April 21, 2025). [52]. ↵ Hart LR . A grocery list learning and memory screening test: Initial performance data stratified by age and IQ . Appl Neuropsychol Adult 2023 ; 30 : 419 – 23 . doi: 10.1080/23279095.2021.1952589 . OpenUrl CrossRef PubMed [53]. ↵ Gong X , Wong PCM , Fung HH , Mok VCT , Kwok TCY , Woo J , et al. The Hong Kong Grocery Shopping Dialog Task (HK-GSDT): A Quick Screening Test for Neurocognitive Disorders . Int J Environ Res Public Health 2022 ; 19 : 13302 . doi: 10.3390/ijerph192013302 . OpenUrl CrossRef PubMed [54]. ↵ Lundberg S , Lee S-I . A Unified Approach to Interpreting Model Predictions 2017 . doi: 10.48550/arXiv.1705.07874 . OpenUrl CrossRef [55]. ↵ Regularization and Variable Selection Via the Elastic Net | Journal of the Royal Statistical Society Series B: Statistical Methodology | Oxford Academic n.d. https://academic.oup.com/jrsssb/article-abstract/67/2/301/7109482?redirectedFrom=fulltext (accessed March 23, 2025). [56]. ↵ Guyon I , Weston J , Barnhill S , Vapnik V . Gene Selection for Cancer Classification using Support Vector Machines . Machine Learning 2002 ; 46 : 389 – 422 . doi: 10.1023/A:1012487302797 . OpenUrl CrossRef [57]. ↵ Lorius N , Locascio JJ , Rentz DM , Johnson KA , Sperling RA , Viswanathan A , et al. Vascular disease and risk factors are associated with cognitive decline in the Alzheimer’s disease spectrum . Alzheimer Dis Assoc Disord 2015 ; 29 : 18 – 25 . doi: 10.1097/WAD.0000000000000043 . OpenUrl CrossRef PubMed [58]. ↵ Dorogush AV , Ershov V , Gulin A . CatBoost: gradient boosting with categorical features support 2018 . doi: 10.48550/arXiv.1810.11363 . OpenUrl CrossRef [59]. Chen T , Guestrin C . XGBoost: A Scalable Tree Boosting System . Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016 , p. 785 – 94 . doi: 10.1145/2939672.2939785 . OpenUrl CrossRef [60]. Ke G , Meng Q , Finley T , Wang T , Chen W , Ma W , et al. LightGBM: A Highly Efficient Gradient Boosting Decision Tree . Advances in Neural Information Processing Systems , vol. 30 , Curran Associates, Inc.; 2017 . [61]. SciPy 1.0: fundamental algorithms for scientific computing in Python | Nature Methods n.d. https://www.nature.com/articles/s41592-019-0686-2 (accessed April 21, 2025). [62]. Pedregosa F , Varoquaux G , Gramfort A , Michel V , Thirion B , Grisel O , et al. Scikit-learn: Machine Learning in Python . Journal of Machine Learning Research 2011 ; 12 : 2825 – 30 . OpenUrl [63]. ↵ Breiman L. Random Forests . Machine Learning 2001 ; 45 : 5 – 32 . doi: 10.1023/A:1010933404324 . OpenUrl CrossRef [64]. ↵ Nasreddine ZS , Phillips NA , Bédirian V , Charbonneau S , Whitehead V , Collin I , et al. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment . J Am Geriatr Soc 2005 ; 53 : 695 – 9 . doi: 10.1111/j.1532-5415.2005.53221.x . OpenUrl CrossRef PubMed Web of Science [65]. ↵ Ruopp MD , Perkins NJ , Whitcomb BW , Schisterman EF . Youden Index and Optimal Cut-Point Estimated from Observations Affected by a Lower Limit of Detection . Biom J 2008 ; 50 : 419 – 30 . doi: 10.1002/bimj.200710415 . OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted May 11, 2025. Download PDF 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 ReCOGnAIze app to detect mild cognitive impairment and vascular cognitive impairment 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 ReCOGnAIze app to detect mild cognitive impairment and vascular cognitive impairment Adnan Azam Mohammed , Ashwati Vipin , Leow Yi Jin , Eliana Setiabudi , Farid Tan , Hitesh Agarwal , Kai Xin Liau , Pricilia Tanoto , Shan Yao Liew , Bocheng Qiu , Gurveen Kaur Sandhu , Jia Dong James Wang , Kiirtaara Aravindhan , Nagaendran Kandiah medRxiv 2025.05.10.25327352; doi: https://doi.org/10.1101/2025.05.10.25327352 Share This Article: Copy Citation Tools ReCOGnAIze app to detect mild cognitive impairment and vascular cognitive impairment Adnan Azam Mohammed , Ashwati Vipin , Leow Yi Jin , Eliana Setiabudi , Farid Tan , Hitesh Agarwal , Kai Xin Liau , Pricilia Tanoto , Shan Yao Liew , Bocheng Qiu , Gurveen Kaur Sandhu , Jia Dong James Wang , Kiirtaara Aravindhan , Nagaendran Kandiah medRxiv 2025.05.10.25327352; doi: https://doi.org/10.1101/2025.05.10.25327352 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 Neurology Subject Areas All Articles Addiction Medicine (567) Allergy and Immunology (863) Anesthesia (295) Cardiovascular Medicine (4408) Dentistry and Oral Medicine (443) Dermatology (380) Emergency Medicine (606) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1505) Epidemiology (15202) Forensic Medicine (30) Gastroenterology (1119) Genetic and Genomic Medicine (6568) Geriatric Medicine (666) Health Economics (994) Health Informatics (4508) Health Policy (1365) Health Systems and Quality Improvement (1608) Hematology (537) HIV/AIDS (1262) Infectious Diseases (except HIV/AIDS) (15902) Intensive Care and Critical Care Medicine (1103) Medical Education (619) Medical Ethics (144) Nephrology (665) Neurology (6572) Nursing (345) Nutrition (998) Obstetrics and Gynecology (1139) Occupational and Environmental Health (954) Oncology (3319) Ophthalmology (967) Orthopedics (369) Otolaryngology (420) Pain Medicine (435) Palliative Medicine (129) Pathology (662) Pediatrics (1686) Pharmacology and Therapeutics (691) Primary Care Research (710) Psychiatry and Clinical Psychology (5420) Public and Global Health (9203) Radiology and Imaging (2190) Rehabilitation Medicine and Physical Therapy (1367) Respiratory Medicine (1191) Rheumatology (593) Sexual and Reproductive Health (709) Sports Medicine (529) Surgery (709) Toxicology (99) Transplantation (288) 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:'9fe69e77a82058f4',t:'MTc3OTIzMDU5MA=='};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())}}}})();