Scholarlink

<p>Computer-aided applications are currently on the rise, particularly in time-sensitive and highly accurate applications such as healthcare. The rate by which dementia cases have been significantly increasing despite clinical reports and psychometric tests being utilized by clinicians to diagnose it illustrates a health challenge and global concern about the techniques being used in the Global medical analysis of the condition. Therefore, with the help of machine learning (ML) algorithms, the aim of this study is to classify patients with dementia using various clinical parameters to help in quick and highly accurate diagnosis. Seven ML models: logistic regression, decision tree, random forest, support vector machine, gradient boosting, voting classifier and stacking classifier, were evaluated using several performance metrics to enhance the efficiency of dementia diagnosis. Interpretation of the model results was based on an explainable algorithm: SHapley Additive exPlanations (SHAP). All classifiers achieved high accuracy in categorising the dementia dataset where random forest, logistic regression and voting classifier achieved high accuracy (92%), high precision (98%), high recall (87% to 88%), high F1-score (92% to 93%) and high AUC (95%) which indicating the classifiers reliability in reducing false positive and false negative dementia cases. The used approach achieves more robustness and easy interpretability compared to other techniques applied in literature through the application of multiple ML techniques, a thorough cross-validation [The ensemble model had statistical significant difference when compared to the single model (AUC: 0.972 vs 0.924, p&lt;0.05), but not with other ensemble models (AUC: 0.972 vs 0.966, p=0.55)], and SHAP analysis. The applied methodology can be used to facilitate the early detection of dementia, enable early treatment interventions, and SHAP-based interpretability can help increase transparency during this decision-making process, thereby increasing their integration into hospital operations. Such an approach can help neurologists and radiologists to effectively screen patients and prioritize further evaluations, which will ultimately lead to a minimization of diagnostic delays.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Applications of Machine Learning Algorithms in Dementia Classification Using Eight Clinical Diagnostic Measures

  • 期刊论文

  • 发表日期：

    2025年12月20日

查看全文
  • Natacha Usanase,
  • Abdullahi G. Usman,
  • Dilber Uzun Ozsahin

摘要

Computer-aided applications are currently on the rise, particularly in time-sensitive and highly accurate applications such as healthcare. The rate by which dementia cases have been significantly increasing despite clinical reports and psychometric tests being utilized by clinicians to diagnose it illustrates a health challenge and global concern about the techniques being used in the Global medical analysis of the condition. Therefore, with the help of machine learning (ML) algorithms, the aim of this study is to classify patients with dementia using various clinical parameters to help in quick and highly accurate diagnosis. Seven ML models: logistic regression, decision tree, random forest, support vector machine, gradient boosting, voting classifier and stacking classifier, were evaluated using several performance metrics to enhance the efficiency of dementia diagnosis. Interpretation of the model results was based on an explainable algorithm: SHapley Additive exPlanations (SHAP). All classifiers achieved high accuracy in categorising the dementia dataset where random forest, logistic regression and voting classifier achieved high accuracy (92%), high precision (98%), high recall (87% to 88%), high F1-score (92% to 93%) and high AUC (95%) which indicating the classifiers reliability in reducing false positive and false negative dementia cases. The used approach achieves more robustness and easy interpretability compared to other techniques applied in literature through the application of multiple ML techniques, a thorough cross-validation [The ensemble model had statistical significant difference when compared to the single model (AUC: 0.972 vs 0.924, p<0.05), but not with other ensemble models (AUC: 0.972 vs 0.966, p=0.55)], and SHAP analysis. The applied methodology can be used to facilitate the early detection of dementia, enable early treatment interventions, and SHAP-based interpretability can help increase transparency during this decision-making process, thereby increasing their integration into hospital operations. Such an approach can help neurologists and radiologists to effectively screen patients and prioritize further evaluations, which will ultimately lead to a minimization of diagnostic delays.