Automatic Spontaneous Speech Analysis for the Detection of Cognitive Functional Decline in Older Adults: Multilanguage Cross-Sectional Study

• DOI: 10.2196/50537
• Publication Date: 2024-02-13T00:22:05Z
• Authors: Emilia Ambrosini; Chiara Giangregorio; Eugenio Lomurno; Sara Moccia; Marios Milis; Christos Loizou; Domenico Azzolino; Matteo Cesari; Manuel Cid Gala; Carmen Galán de Isla; Jonathan Gomez-Raja; Nunzio Alberto Bo...; Matteo Matteucci; Simona Ferrante
• Keywords: Cognitive Decline; Binary classification

Background The rise in life expectancy is associated with an increase long-term and gradual cognitive decline. Treatment effectiveness enhanced at the early stage of disease. Therefore, there a need to find low-cost ecological solutions for mass screening community-dwelling older adults. Objective This work aims exploit automatic analysis free speech identify signs function Methods A sample 266 participants than 65 years were recruited Italy Spain divided into 3 groups according their Mini-Mental Status Examination (MMSE) scores. People asked tell story describe picture, voice recordings used extract high-level features on different time scales automatically. Based these features, machine learning algorithms trained solve binary multiclass classification problems by using both mono- cross-lingual approaches. enriched Shapley Additive Explanations model explainability. Results In Italian data set, healthy (MMSE score≥27) automatically discriminated from mildly impaired (20≤MMSE score≤26) those moderate severe impairment (11≤MMSE score≤19) accuracy 80% 86%, respectively. Slightly lower performance was achieved Spanish multilanguage sets. Conclusions proposes transparent unobtrusive assessment method, which might be included mobile app large-scale monitoring functionality Voice confirmed important biomarker decline due its noninvasive easily accessible nature.