A Machine Learning Approach for Early Diagnosis of Cognitive Impairment Using Population-Based Data

Abstract

Background: The major mechanisms of dementia and cognitive impairment are vascular and neurodegenerative processes. Early diagnosis of cognitive impairment can facilitate timely interventions to mitigate progression. Objective: This study aims to develop a reliable machine learning (ML) model using socio-demographics, vascular risk factors, and structural neuroimaging markers for early diagnosis of cognitive impairment in a multi-ethnic Asian population. Methods: The study consisted of 911 participants from the Epidemiology of Dementia in Singapore study (aged 60– 88 years, 49.6% male). Three ML classifiers, logistic regression, support vector machine, and gradient boosting machine, were developed. Prediction results of independent classifiers were combined in a final ensemble model. Model performances were evaluated on test data using F1 score and area under the receiver operating curve (AUC) methods. Post modelling, SHapely Additive exPlanation (SHAP) was applied on the prediction results to identify the predictors that contribute most to the cognitive impairment prediction. Findings: The final ensemble model achieved a F1 score and AUC of 0.87 and 0.80 respectively. Accuracy (0.83), sensitivity (0.86), specificity (0.74) and predictive values (positive 0.88 negative 0.72) of the ensemble model were higher compared to the independent classifiers. Age, ethnicity, highest education attainment and neuroimaging markers were identified as important predictors of cognitive impairment. Conclusion: This study demonstrates the feasibility of using ML tools to integrate multiple domains of data for reliable diagnosis of early cognitive impairment. The ML model uses easy-to-obtain variables and is scalable for screening individuals with a high risk of developing dementia in a population-based setting.

Predicting early Alzheimer’s with blood biomarkers and clinical features

“…Zhu et al 32 used MRI data and the APOE4 genotype in a wide neural network to predict cognitive decline in A -positive individuals with an accuracy of 0.86. An ensemble of logistic regression, support vector machine, and gradient boosting methods achieved an AUC of 0.87 for early diagnosis of cognitive impairment using demographic and MRI data from the Epidemiology of Dementia in Singapore study 30 . A neural network for predicting MCI diagnosis using radiomic features and amyloid brain PET attained an AUC of 0.90 using 656 subjects from ADNI and a EudraCT (European Union Drug Regulating Authorities Clinical Trials Database) cohort 31 .…”

Application of machine learning in measurement of ageing and geriatric diseases: a systematic review

“…The maximally selected rank statistics RF outperformed the other random forest models. In case of feature selection, RF min depth filter produced most accurate models NA 40 Tan et al [ 55 ] 2023 To develop a reliable ML model using socio-demographics, vascular risk factors, and structural neuroimaging markers for early diagnosis of cognitive impairement in multi-ethnic Asian population 911 participants from Epidemiology of Dementia in Singapore study Neurodegenrative Disease ML models- logistic regression, support vector machine, gradient boosting machine voting ensemble- SHAP According to the voting ensemble, the important predictors of cognitive impairement are age, ethnicity, education attainment, and structural neuroimaging LR (accuracy = 0.71, F1 = 0.78, AUC = 0.69, FPR = 0.38, sensitivity = 0.75, specificity = 0.62, PPV = 0.81, NPV = 0.54), SVM (accuracy = 0.74, F1 = 0.81, AUC = 0.71, FPR = 0.40, sensitivity = 0.81, specificity = 0.60, PPV = 0.81, NPV = 0.59), GBM (accuracy = 0.73, F1 = 0.79, AUC = 0.73, FPR = 0.29, sensitivity = 0.74, specificity = 0.71, PPV = 0.85, NPV = 0.56), Ensemble (accuracy = 0.83, F1 = 0.87, AUC = 0.80, FPR = 0.26, sensitivity = 0.86, specificity = 0.74, PPV = 0.88, NPV = 0.72) 41 Hu et al [ 56 ] 2021 To build a prediction model based on ML for cognitive impairement among Chinese community dwelling elderly people with normal cognition 6718 individuals of age > 60, with MMSE score > = 18, not having any severe disease from the Chinese Longitudinal Health Longevity Survey (CLHLS) Neurodegenrative Disease To access 3-year risk of developing cognitive impairement, Ml models used- Random forest, XGBoost, Naïve Bayes, Logistic regression A nomogram was established to vividly present the prediction model Features selected to develop model- age, instrumental activities of daily living, marital status, and baseline cognitive function Older people with nomogram score less than 170 are considered to have a low 3-year risk, and more than 173 are considered at higher risk AUC, optimal cut off, sensitivity, specificity, accuracy, specificity/sensitivity values were reported for logisitc regression, random forest, naïve bayes, XG Boost both for validation dataset and test dataset 42 Fukunishi et al [ 57 ] 2020 To predict the risk of Alzheimer-type dementia for persons aged over 78 without receiving long-term care services using regularly collected claim data 48,123 persons from claim data of health insurance and long-term care insurance in Japan Ne...…”

A risk prediction model based on machine learning for early cognitive impairment in hypertension: Development and validation study

“…Kang et al ( 60 ) developed and validated the Aβ positive predictive model for amnestic mild cognitive impairment (aMCI) using two-stage modeling based on machine learning with good accuracy (AUC: 0.892). Tan et al ( 6 ) used three classifiers (logistic regression, support vector machine, and gradient enhancer) to construct a set model for predicting cognitive impairment, with F1 score (0.87), AUC (0.80), accuracy (0.83), sensitivity (0.86), and specificity (0.74). In this study, we used three classifiers (LR, XGB, and GBN) to develop machine learning predictive models for early cognitive impairment in hypertension for the first time and obtained stable predictive performance.…”

“…AUC, accuracy, specificity, sensitivity, and F1 scores were used to evaluate the performance of the machine learning models. The classification confusion matrix definition is that individuals with cognitive decline are considered true positive (TP) and true negative (TN) if they are accurately predicted by the machine learning model; In contrast, it is considered false positive (FP) or false negative (FN) ( 6 ). AUC, the area under the ROC curve, the larger the value, the better the classification effect.…”

“…The traditional approach to the diagnosis of cognitive impairment in hypertension focuses on cognitive and neuropsychological assessment but is often criticized for its lag ( 6 ). More recently, although amyloid proteins, tau proteins, and several structural magnetic resonance imaging (MRI) indicators have been recognized as promising pathologic markers ( 7 , 8 ), high costs and complex inspection procedures still limit their widespread use.…”