MemoryTELL - speech biomarkers to support dementia diagnosis

  • Research type

    Research Study

  • Full title

    Pilot Study - Validation of the MemoryTELL app that uses speech and language digital biomarkers to support expedited dementia diagnoses in clinical practice.

  • IRAS ID

    351110

  • Contact name

    Aaron Peace

  • Contact email

    aaron.peace@westerntrust.hscni.net

  • Sponsor organisation

    MemoryTELL

  • Duration of Study in the UK

    1 years, 0 months, 1 days

  • Research summary

    This study aims to evaluate and validate an app for expediting the dementia diagnosis pathway through analysing speech and language biomarkers based on patient voice recordings. This pilot project requires the recruitment of a small cohort of individuals (n=60), with capacity to provide consent, referred to attend a WHSCT memory clinic for assessment and potential diagnosis. The application (web based but downloadable module to use off-line) is intended to provide an efficient, accessible method for diagnosing dementia, with a focus on differentiating between disease phenotypes, predicting symptom severity, and identifying potential brain atrophy patterns before clinical diagnosis.

    Participants will provide informed consent and complete a series of speech-based tasks at their research appointment, with voice recordings capturing both acoustic ('prosodic') and linguistic ('semantic') features. These recordings will then be analysed using Automated Language and Speech Analysis (ALSA) tools.
    Alongside the voice recordings, participants will complete a questionnaire to collect relevant health and lifestyle information from participants. Tasks and data will be anonymised to ensure participant confidentiality, with all data securely stored in compliance with GDPR. Participants will be fully informed that voice recordings can be classed as identifiable, as there is a possibility of being recognised based on voice alone. The MemoryTELL app analyses the recordings to extract metrics such as speech speed, average duration of syllables, average duration of breaks, and emotional magnitude, which have been shown to elicit distinguishing verbal behaviours in individuals diagnosed with dementia, including Alzheimer's disease and other subtypes.

    This study will compare the app's results with standard clinical assessments, allowing for an evaluation of its potential as an adjunct diagnostic tool in clinical practice. The primary research goal is to assess the app's ability to support early dementia diagnosis through speech analysis, providing clinicians with an innovative, non-invasive method to aid in the existing diagnostic pathway.

  • REC name

    Wales REC 6

  • REC reference

    25/WA/0096

  • Date of REC Opinion

    24 Mar 2025

  • REC opinion

    Favourable Opinion