Dementia and Heart Failure Classification Using Optimized Weighted Objective Distance and Blood Biomarker-Based...

Area of Science:

• Biomedical Engineering
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Dementia and heart failure are significant global health challenges, particularly with aging populations.
• Accurate diagnosis is hindered by limited access to advanced diagnostic equipment and tests.
• Distinguishing between dementia and heart failure is crucial to prevent misdiagnosis and inappropriate patient referrals.

Purpose of the Study:

• To introduce a novel measurement, the optimized weighted objective distance (OWOD), for improved classification of dementia and heart failure.
• To enhance model generalization and classification performance using a data-driven approach.
• To integrate novel blood biomarker features with existing risk factors for robust diagnostic capabilities.

Main Methods:

• Developed the optimized weighted objective distance (OWOD), a modified weighted objective distance metric.
• Applied multi-feature distance normalization and identified key classification features.
• Integrated 20 features, including risk factors and proposed blood biomarkers, from 10,000 electronic health records.
• Utilized machine learning models for classification and performance evaluation.

Main Results:

• The OWOD-based classification method achieved high performance metrics: 95.45% accuracy, 96.14% precision, 94.70% recall, and 95.42% F1-score.
• The area under the ROC curve reached 97.10%, indicating strong discriminative power.
• The OWOD method outperformed other machine learning models, including gradient boosting, decision tree, neural network, and support vector machine.

Conclusions:

• The optimized weighted objective distance (OWOD) offers a powerful and accurate tool for differentiating between dementia and heart failure.
• The integration of blood biomarkers significantly enhances classification performance.
• This novel approach holds promise for improving diagnostic accuracy in clinical settings, especially where resources are limited.