Eli Lilly Licenses ADDF’s SpeechDx Dataset for Alzheimer’s Research
Eli Lilly joins global partners to leverage longitudinal speech data for early Alzheimer’s disease detection.
1Strategic Data Licensing Agreement
On October 28, 2026, Eli Lilly and Company officially licensed the SpeechDx dataset from the Alzheimer’s Drug Discovery Foundation. This resource, developed by the foundation’s Diagnostics Accelerator, provides longitudinal speech and biomarker data. Lilly becomes the third organization to access this collection, following previous partnerships with Siemens Healthineers and the company Callyope.
2Defining the SpeechDx Resource
SpeechDx is currently recognized as the world’s largest harmonized longitudinal collection of speech and biomarker data. The dataset tracks the Alzheimer’s continuum, ranging from cognitively unimpaired individuals to those with advanced dementia. It aims to provide researchers with the comprehensive information necessary to develop and validate new digital markers for clinical use.
3Clinical Scope and Methodology
The observational study encompasses 2,650 participants across global sites, including Boston University and the Barcelona Beta Brain Research Centre. Participants provide speech data quarterly via a dedicated app, performing tasks like picture description and story recall. This audio is paired with neuropsychological testing, MRI imaging, and blood-based amyloid status to ensure high-quality validation.
4Advancing Early Disease Detection
Eli Lilly intends to use this data to make Alzheimer’s diagnosis more accurate, accessible, and less invasive. Mark Mintun, group vice president at Lilly, noted that mild symptoms are often mistaken for normal aging, which delays critical interventions. By integrating speech measurements with established biomarkers, the company seeks to identify cognitive decline at its earliest stages.
5Scalability of Digital Biomarkers
The foundation highlights that speech can be captured using standard smartphone or computer microphones, offering a highly scalable monitoring tool. By training AI models on this gold-standard dataset, researchers hope to transform speech into a dependable digital instrument. This approach addresses the critical gap in tracking how individual speech patterns change over time.
6Data Governance and Security
To protect participant privacy, all audio is encrypted and pseudonymized before entering the AD Curation Studio. Researchers work within a controlled environment where data cannot be downloaded or exported. A dedicated committee reviews every proposed use of the dataset to ensure strict compliance with all ethical and legal requirements for sensitive health information.
7Accelerator Funding and Evolution
The Diagnostics Accelerator, launched in 2018, has invested $100 million across 76 global programs. A third phase announced in March 2026 brought total funding to $150 million, supported by partners like Biogen and the Dolby family. This initiative previously enabled the first FDA-approved Alzheimer’s blood test, demonstrating the program's impact on diagnostic innovation.
8Future Research Trajectory
The SpeechDx study is scheduled to conclude in Q4 2027, with data releases continuing in quarterly batches. As partners like Callyope and Siemens Healthineers utilize the data for AI-foundation models, the project moves closer to establishing speech as a validated clinical instrument. This collaborative effort marks a significant shift toward data-driven, non-invasive neurodegenerative disease management.
Original article: Unite.AI — Eli Lilly Licenses ADDF’s SpeechDx Dataset for Alzheimer’s Research