A Novel Approach to Early Detection of Alzheimer’s Disease Using Handwriting Analysis with Quadratic Discriminant Analysis
Yukitaka Watanabe
Alhambra High School, Martinez, California
Publication date: January 30, 2025
Alhambra High School, Martinez, California
Publication date: January 30, 2025
DOI: http://doi.org/10.34614/JIYRC202434
ABSTRACT
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that impacts millions worldwide, with cases expected to rise as the global population ages. Early diagnosis is crucial for slowing disease progression, but conventional diagnostic methods, such as cerebrospinal fluid analysis and magnetic resonance imaging, are invasive and costly, limiting their accessibility and effectiveness in widespread early detection efforts. This study explores handwriting analysis as a non-invasive, cost-effective diagnostic tool for Alzheimer’s, leveraging unique motor and cognitive signatures associated with the disease. Using the DARWIN dataset, which comprises handwriting samples from Alzheimer’s patients and healthy controls, I applied several machine learning models—including LGBM Classifier, Support Vector Machines, Quadratic Discriminant Analysis, and Random Forest classifiers—to distinguish between these groups. To improve model performance, I conducted Principal Component Analysis (PCA) as a preprocessing step, reducing dimensionality and capturing the most informative features, including tremor frequency, pressure consistency, and movement speed. Quadratic Discriminant Analysis (QDA) was identified as the most effective model, achieving high classification accuracy. These features correlate with specific handwriting tasks and reveal distinct cognitive and motor control patterns affected by Alzheimer’s. Our findings suggest that handwriting analysis, when combined with machine learning, offers a promising, accessible avenue for early Alzheimer’s diagnosis, supporting interventions that can potentially mitigate cognitive decline. Future research should explore integrating this model with other non-invasive diagnostic tools to further enhance early detection capabilities.
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that impacts millions worldwide, with cases expected to rise as the global population ages. Early diagnosis is crucial for slowing disease progression, but conventional diagnostic methods, such as cerebrospinal fluid analysis and magnetic resonance imaging, are invasive and costly, limiting their accessibility and effectiveness in widespread early detection efforts. This study explores handwriting analysis as a non-invasive, cost-effective diagnostic tool for Alzheimer’s, leveraging unique motor and cognitive signatures associated with the disease. Using the DARWIN dataset, which comprises handwriting samples from Alzheimer’s patients and healthy controls, I applied several machine learning models—including LGBM Classifier, Support Vector Machines, Quadratic Discriminant Analysis, and Random Forest classifiers—to distinguish between these groups. To improve model performance, I conducted Principal Component Analysis (PCA) as a preprocessing step, reducing dimensionality and capturing the most informative features, including tremor frequency, pressure consistency, and movement speed. Quadratic Discriminant Analysis (QDA) was identified as the most effective model, achieving high classification accuracy. These features correlate with specific handwriting tasks and reveal distinct cognitive and motor control patterns affected by Alzheimer’s. Our findings suggest that handwriting analysis, when combined with machine learning, offers a promising, accessible avenue for early Alzheimer’s diagnosis, supporting interventions that can potentially mitigate cognitive decline. Future research should explore integrating this model with other non-invasive diagnostic tools to further enhance early detection capabilities.