AI-Powered Speech Assessment Detects Alzheimer's Pathology Through Automated Phone Calls
Key Insights
ki:elements (search)' Speech Biomarker for Cognition (SB-C) successfully detected cognitive impairment and Alzheimer's disease (search) pathology through automated phone calls in a study of 736 participants across five European cohorts.
The AI voice agent demonstrated significant diagnostic accuracy, identifying amyloid-beta (search) positivity with an AUC of up to 0.74 and phosphorylated tau 181 (search) positivity with an AUC of up to 0.82.
The 10-minute multilingual assessment offers pharmaceutical companies a validated pre-screening tool to identify biomarker-positive candidates before expensive invasive procedures like lumbar punctures or PET scans.
ki:elements (search) and the PROSPECT-AD consortium (search) have published peer-reviewed research demonstrating that their Speech Biomarker for Cognition (SB-C) can reliably detect cognitive impairment and signal underlying Alzheimer's disease (search) pathology through an AI voice agent that calls participants at home. The study, involving 736 participants across five independent cohorts in Spain, the UK, Germany and Sweden, represents one of the largest and most geographically diverse validations of a speech-based digital cognitive assessment to date.
Diagnostic Performance Across Multiple Biomarkers
The research confirms that the SB-C consistently separated cognitive unimpaired individuals from those with mild cognitive impairment (search) (MCI) or early dementia (search). The tool demonstrated significant ability to classify Alzheimer's CSF biomarker status remotely, identifying amyloid-beta (search) positivity with an AUC of up to 0.74 and phosphorylated tau 181 (search) positivity with an AUC of up to 0.82.
The SB-C proved effective across Spanish, Catalan, German, English, and Swedish language groups, utilizing 70 distinct speech features to index cognitive efficiency. The platform demonstrates strong convergent validity with clinical "gold standards" such as MMSE and PACC-5, providing a window into cognitive health that complements traditional paper tests by quantifying subtle functional speech changes.
Addressing Clinical Trial Recruitment Challenges
Recruiting for Alzheimer's disease (search) clinical trials traditionally represents one of the costliest phases of drug development, often requiring invasive and expensive procedures like lumbar punctures or PET scans to confirm eligibility. The SB-C offers the biopharma industry a validated first-pass enrichment layer, identifying individuals most likely to meet biomarker-positive eligibility criteria before committing to expensive invasive procedures.
The ki:elements (search) Mili platform provides fully automated delivery, calling participants and administering the 10-minute assessment through a voice agent before generating an automatic biomarker score. The SB-C is validated following the Digital Medicine Society's V3 Framework and is currently deployed as a pre-screening tool in multiple Alzheimer's disease (search) clinical trials.
Clinical Impact and Future Applications
"We've known for years that one of the biggest hurdles in prevention trials is getting the right people into studies efficiently and at scale," said Prof Craig Ritchie, Professor of Brain health and Neurodegenerative Medicine at University of St Andrews and CEO and Founder of Scottish Brain Sciences (search), and co-author of the publication. "In this era of early-intervention Alzheimer's research, a validated digital cognitive assessment that correlates with AD pathology in multiple languages, without a clinician, opens up a generation of prevention trials that weren't feasible before."
Dr. Alexandra König, Chief Clinical Research Officer of ki:elements (search) and corresponding author of the study, emphasized the practical implications: "This study gives us real-world, multi-site evidence that a speech assessment can track with established cognitive measures and distinguish patients by their underlying pathology. For sponsors, this translates directly into faster, smarter recruitment, and a lower-burden tool for participants."
The technology represents a significant advancement in decentralized clinical trial execution, offering pharmaceutical companies and research institutions a scalable solution for early detection and patient identification in Alzheimer's disease (search) and other central nervous system conditions.