Related Experiment Video
Updated: Aug 9, 2025
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.6K
Early Prediction of Dementia Using Feature Extraction Battery (FEB) and Optimized Support Vector Machine (SVM) for
Ashir Javeed1,2, Ana Luiza Dallora2, Johan Sanmartin Berglund2
1Aging Research Center, Karolinska Institutet, 171 65 Stockholm, Sweden.
Biomedicines
|February 25, 2023
Summary
This study introduces a novel machine learning model for early dementia prediction, significantly improving accuracy and reducing bias. The developed Feature Extraction Battery-Support Vector Machine (FEB-SVM) model achieved 93.92% accuracy in testing.
Area of Science:
- Neuroscience
- Computer Science
- Artificial Intelligence
Background:
- Dementia is a prevalent cognitive disorder in older adults with no current cure or prevention.
- Dementia symptoms can manifest up to a decade before clinical diagnosis, highlighting the need for early detection.
- Existing machine learning models for dementia prediction suffer from limitations like low accuracy and inherent bias.
Purpose of the Study:
- To address the limitations of current machine learning models for dementia prediction.
- To develop a more accurate and less biased model for early dementia detection.
- To introduce novel feature extraction techniques and optimize existing classification algorithms.
Main Methods:
- Adaptive Synthetic Sampling (ADASYN) technique was employed to mitigate bias in machine learning models.
- Novel feature extraction techniques, termed Feature Extraction Battery (FEB), were developed.
- An optimized Support Vector Machine (SVM) model using a radical basis function (rbf) kernel was proposed for disease classification.
- Support Vector Machine (SVM) hyperparameters were fine-tuned using a grid search approach.
Main Results:
- The proposed FEB-SVM model demonstrated a 6% improvement in dementia prediction accuracy compared to conventional SVM.
- The FEB-SVM model achieved a training accuracy of 98.28% and a testing accuracy of 93.92%.
- Performance metrics included 91.80% precision, 86.59% recall, 89.12% F1-score, and a Matthew's correlation coefficient (MCC) of 0.4987.
- The FEB-SVM model outperformed 12 other state-of-the-art machine learning models in dementia prediction.
Conclusions:
- The developed FEB-SVM model offers a significant advancement in the accuracy and reliability of early dementia prediction.
- The combination of ADASYN, FEB, and optimized SVM provides a robust framework for tackling bias and improving performance in dementia detection.
- This approach holds promise for earlier and more effective interventions in dementia care.
