Abstract
BACKGROUND: Neuropsychiatric symptoms (NPS) can precede cognitive decline in Alzheimer's Disease and serve as prognostic factors of disease progression. Assessing NPS in clinical practice is challenging due to time constraints, subjectivity and limited adaptability of tools to early‐stage cognitive decline. Advancements in automatic speech analysis may enable objective characterization of NPS, while structural brain morphometry associated with speech features offers insights into the underlying mechanisms of NPS. This study explored associations between extracted speech features, gold‐standard NPS assessments and volumetric brain measures. METHOD: Within the PROSPECT‐AD project, data were obtained from the German DELCODE and DESCRIBE cohorts. Analysis included N = 30 healthy controls and N = 44 participants with Subjective Cognitive Decline (SCD)/Mild Cognitive Impairment (MCI) with NPS, assessed by the Geriatric Depression Scale and the Neuropsychiatric Inventory. Participants answered a free‐speech question (“Can you describe a positive event in your life?”). Acoustic features (spectral/temporal/frequency/energy variables) and linguistic (i.e. lexical richness, syntactic complexity) were automatically extracted. Residuals from regression models (adjusted for age, sex, MMSE) were used to compute Spearman rank correlations between speech features and clinical scores measuring depression, apathy, anxiety and agitation. In the SCD/MCI group with NPS (N = 21), adjusted correlations were computed between speech features and baseline volumetric measures of regions of interest (ROI). Hypothesis‐driven mediation analysis between speech features, clinical scales and ROI volumes is ongoing. RESULTS: We found significant associations (p < 0.05) between the severity of NPS and acoustic/linguistic markers, though these did not survive multiple comparisons corrections (Figure 1). For example, anxiety positively correlated with the sum and duration of pauses, while depression positively correlated with vocal tremor. In the SCD/MCI group with NPS, speech features correlated with volumes in key ROIs, especially related to emotion regulation (i.e. amygdala, insular cortex) (Figure 2). CONCLUSION: Our exploratory findings indicate that (mainly acoustic) markers derived from free speech are associated with the severity of NPS as measured by clinical scales and with regional brain volumes in participants with NPS. These dual associations highlight the potential of speech analysis as a non‐invasive, objective tool to assess NPS in early‐stage cognitive decline.