15 April 2025 in Journal of Alzheimer’s Disease
Background: Mild cognitive impairment (MCI), a prodromal stage of Alzheimer's disease (AD) for many individuals, is accompanied by widespread connected speech (CS) changes (e.g., shorter CS samples, mention of fewer semantic content units, lower syntactic complexity). Nevertheless, findings on CS in MCI are heterogeneous. This heterogeneity, combined with the heterogeneity in cognition in MCI suggests that there could exist more than one CS profile in this population. Objective: We aimed to determine if there are multiple CS profiles in MCI and whether these potential CS profiles are characterized by distinct cognitive presentations. Methods: CS characteristics were extracted from the samples of 109 controls and 210 individuals with MCI from the COMPASS-ND study database. A Two-Step Cluster Analysis was then carried out to identify potential CS profiles in MCI. These profiles were compared to one another and to controls in terms of their linguistic and cognitive characteristics. Results: We identified two CS profiles in MCI, characterized by reduced syntactic complexity and semantic content and by dysfluencies, longer CS samples, and reduced semantic idea density and efficiency, respectively. The reduced semantic content/syntactic complexity profile was also characterized by various cognitive difficulties (e.g., visuospatial, episodic memory, executive functioning domains) in comparison with controls, whereas the increased production and reduced idea transmission effectiveness profile had relatively isolated episodic memory difficulties. Conclusions: CS analysis could be a helpful screening tool to identify individuals with MCI who show greater cognitive difficulties and who would most benefit from more extensive cognitive and/or medical testing as well as from cognitive and/or psychological interventions.
26 February 2026 in medRxiv
INTRODUCTION: Connected speech analyses can help characterize linguistic impairments in primary progressive aphasia (PPA) and classify variants, however, manual transcription of speech samples is time-consuming and expensive. Automated speech recognition (ASR) may be efficacious for transcribing PPA speech. METHODS: Transcripts of picture descriptions (109 PPA, 32 healthy controls (HC)) were generated using a manual, automated (Whisper) or semi-automated approach including a quality control (QC) step. We evaluated transcript accuracy, the reliability of ASR-derived linguistic features, and classification performance. RESULTS: Whisper demonstrated lowest error rates for HC, followed by semantic, logopenic and non-fluent PPA variants. Errors correlated with overall disease severity for semantic and logopenic variants. QC of Whisper outputs reduced errors and improved the reliability of linguistic features. Overall, ASR-derived features achieved better classification performance than manual transcription features. DISCUSSION: Results support the use of off-the-shelf ASR for scalable, cost-efficient transcription of PPA speech and classification.