A speech based diagnostic method for Alzheimer disease using machine learning

Alzheimer’s Disease (AD), the most common form of dementia, is a neurodegenerative condition which evolves over time. Patients experience short-term memory lapses at the start of this process, and by the end, they depend exclusively on signs. While there is no cure for AD, Medical treatments can be pursued for premature patients. Individuals suffering from early-stage Alzheimer’s disease have signs of linguistic impairment, including issues with word recall and word seeking. In this paper, we use stacking, an ensemble Machine Learning algorithm that learns how to combine the best predictions from multiple high-performing Machine Learning models, such as the K-Nearest Neighbors algorithm, Random Forest algorithm, Decision Tree algorithm, Support Vector Machine algorithm and Multi-Layer Perceptron algorithm, to accurately predict AD. The data-set is gathered from the dementia Pitt corpus repository. Our methodology focuses on utilizing extracted features from speech signals to detect abnormalities and deduce whether a patient has AD. 78% of AD sufferers and Healthy participants can be distinguished using the proposed technique. The average test accuracy using stacking classifier is 64%. Machine Learning algorithms employ 13 dimensionality reduction technique and an average accuracy reported by Neighborhood Component Analysis is 73%. 97% accuracy was attained using stacking classifier in machine learning algorithms. This data set demonstrates the ability to detect AD using only audio information. © 2023, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.