ki:elements

Assessing neuropsychiatric symptoms through speech analysis in early dementia patients

Zampeta-Sofia Alexopoulou, Elisa Mallick, Johannes Tröger, Daphne ter Huurne, Nina Possemis, Inez Ramakers & Alexandra König

* Talk presented at the 34th Alzheimer Europe Conference, Geneva (Switzerland)

Abstract

Introduction: Neuropsychiatric symptoms (NPS) hold prognostic value for cognitive decline in Alzheimer’s Disease (AD). However, they are often inadequately addressed in clinical settings. Speech features are impacted in various NPS like depression, anxiety or apathy, even in early cognitive decline. Subtle speech alterations might predict not only cognitive but also behavioral changes in AD, aiding in diagnosis and disease monitoring. This study aimed to examine associations between automatically extracted speech and language features and traditional assessment measurements of NPS.

Method: Within the DeepSpa project, speech was recorded from N = 38 patients with Subjective Cognitive Decline, N = 20 patients with Mild Cognitive Impairment and N = 3 dementia patients, who were included in the BioBank Alzheimer Centre Limburg study. Participants answered a narrative question (“What did you do during your last vacation?”), while NPS were assessed using the Geriatric Depression Scale and the Neuropsychiatric Inventory. Acoustic and linguistic features were automatically extracted. For each feature, a linear regression model was fitted, using age, sex and MMSE as predictors. The model’s residuals were used to compute Spearman rank correlations with the clinical scales.

Results: Acoustic and linguistic features were associated with agitation, apathy, anxiety and depression. Notably, agitation significantly correlated with Brunet’s index (metric of lexical richness) (r = 0.26, p = 0.04), number of pauses (r = 0.31, p = 0.02) and mean pitch (r = -0.36, p <0.001). Depression significantly correlated with Honore’s statistic (metric of lexical richness) (r = -0.36, p <0.001), local absolute jitter (r = 0.35, p = 0.01) and mean local shimmer (r = 0.23, p = 0.07).

Conclusion: Alterations in free speech patterns appear linked to particular NPS, which are considered dementia risk factors and are focal points in clinical trials. Utilizing objective assessment tools like speech analysis could improve NPS management and treatment.

Share this article