Identificación de la enfermedad de Alzheimer a través del habla: un enfoque multilingüe

Resumen

Alzheimer's disease, the leading form of dementia among elderly individuals worldwide, has significant social and economic repercussions. It is characterized by memory loss and changes in language, cognition, and emotions, irreversibly affecting neurons. Early diagnosis is crucial but challenging, as it relies on detailed medical evaluations, cognitive tests, and complex exams that are often expensive and inaccessible, particularly for low-income individuals. In this context, advanced computational techniques, such as machine learning (ML), emerge as promising non-invasive alternatives for the early detection of the disease. This study introduces a multilingual ML-based approach focusing on paralinguistic and emotional speech characteristics as biomarkers for Alzheimer's identification. The experiments yielded results with accuracies reaching 81% for English and 87.50% for Portuguese. Additionally, integrating this methodology with the state-of-the-art model by Haider, Fuente, and Luz(1) resulted in an average accuracy of 81.70%, surpassing their original results.