Speech, language, p-tau217 and neurofilament light chain in an Alzheimer's Disease at risk population

IntroductionAlzheimers disease (AD) affects speech, language, and executive functions. Combining blood biomarkers with non-invasive speech analysis may aid early detection. MethodsA speech battery was administered to three groups aged 65-70: cognitively normal low-risk (LRC), high-risk (HRC), and mild cognitive impairment (MCI). ResultsVariance in Mel-Frequency Cepstral Coefficients (MFCCs) predicted p-tau217 (F = 3.71, p = 0.005), with group-specific effects in LRC and HRC. In MCI, articulation rate, speech percentage, and pause percentage predicted p-tau217 (all p < 0.01). Language features including idea density, content density, and open-class word rate also predicted p-tau217 (all p < 0.05). MFCCs and syllable timing correlated with NfL (all p < 0.01). Lexical diversity differed between groups, notably in HRC and MCI. DiscussionSpeech timing, fluency, and voice quality, along with lexical richness and idea density, predicted p-tau217 and NfL, supporting acoustic and linguistic features as non-invasive digital biomarkers for early AD-related neurodegeneration.