Machine Learning-based Classification of Alzheimer’s Disease Given Clinical and Demographic Data
Christine Johansen
Dominican Academy, New York City, USA
Publication date: November 20, 2025
Dominican Academy, New York City, USA
Publication date: November 20, 2025
DOI: http://doi.org/10.34614/JIYRC2025II40
ABSTRACT
Alzheimer’s disease (AD) is an extremely prevalent and incurable neurodegenerative disorder. AD diagnosis is often difficult due to challenges in recognizing early symptoms, leading to the use of machine learning-based AD detection. To improve diagnostic accuracy without dependence on MRI technology, as they can be too expensive or otherwise inaccessible, we programmed a logistic regression model and 36 deep neural networks to classify subjects as with dementia or without dementia given clinical and demographic data. Both of these models were tested on 71 samples (total samples n=354). The logistic regression model achieved an accuracy of 95.77% while the 36 neural networks achieved a maximum accuracy of 100% and an average accuracy of 96.38%. All of these models outperformed traditional AD diagnostic tests, including the Montreal Cognitive Assessment, suggesting that a combination of traditional tests, physician input, and machine-learning algorithms will lead to more accurate Alzheimer’s diagnosis.
Alzheimer’s disease (AD) is an extremely prevalent and incurable neurodegenerative disorder. AD diagnosis is often difficult due to challenges in recognizing early symptoms, leading to the use of machine learning-based AD detection. To improve diagnostic accuracy without dependence on MRI technology, as they can be too expensive or otherwise inaccessible, we programmed a logistic regression model and 36 deep neural networks to classify subjects as with dementia or without dementia given clinical and demographic data. Both of these models were tested on 71 samples (total samples n=354). The logistic regression model achieved an accuracy of 95.77% while the 36 neural networks achieved a maximum accuracy of 100% and an average accuracy of 96.38%. All of these models outperformed traditional AD diagnostic tests, including the Montreal Cognitive Assessment, suggesting that a combination of traditional tests, physician input, and machine-learning algorithms will lead to more accurate Alzheimer’s diagnosis.
