Alzheimer’s Drug Discovery Foundation’s Diagnostics Accelerator (DxA) Partners with Siemens Healthineers to License DxA’s SpeechDx Dataset, Gold-Standard Dataset for Alzheimer’s Speech Biomarkers

The Alzheimer’s Drug Discovery Foundation (ADDF)’s Diagnostics Accelerator (DxA) today announced a new partnership with Siemens Healthineers to license the DxA’s SpeechDx dataset. SpeechDx, DxA’s most sizeable investment to date, was launched in 2023 as a study aimed at creating the largest longitudinal dataset consisting of harmonized speech and biomarker data to help accelerate the detection, diagnosis and monitoring of Alzheimer’s disease. Under the agreement, Siemens will license the SpeechDx dataset to develop speech-based biomarkers for Alzheimer’s disease. 

“We have long known that Alzheimer’s pathology can appear 20-30 years before symptoms arise, creating an opportunity to detect and intervene earlier on in the disease progression,” said Howard Fillit, MD, Co-Founder and Chief Science Officer at the ADDF. “Previous research suggests subtle changes in speech can be one of the earliest indicators of cognitive decline in Alzheimer’s, and our partnership with Siemens represents the first of many collaborations that will utilize SpeechDx dataset to identify a new class of speech-based biomarkers and digital tools for Alzheimer’s.” 

This cutting-edge initiative provides researchers with rigorously collected longitudinal speech, clinical, and biomarker data – allowing scientists to identify speech-based biomarkers and develop novel diagnostic and prognostic tools for Alzheimer’s. 

"The earlier we identify the prospect of life-altering degenerative diseases, the greater impact we can potentially have on patients and their families' lives. Predicting Alzheimer's disease at the earliest possible stage with speech analysis and laboratory testing is untapped potential. The limitation to achieving this milestone has been access to large cohorts of ethically sourced patient data that might provide clues about how early we may be able to predict patients' degeneration. Through this collaborative agreement, we can explore what opportunities may be possible for earlier detection," said Arejas J. Uzgiris, PhD, neurology innovation lead and head of the Center for Innovation in Diagnostics for North America at Siemens Healthineers. 

SpeechDx serves as the gold-standard dataset for developing diagnostic and prognostic speech-based biomarkers that may signal Alzheimer’s at its earliest stages. The DxA recognized a critical gap in the data fueling the high-potential field of speech diagnostics for Alzheimer’s, establishing SpeechDx as a unified speech dataset for industry, biotech, and academic leaders to accelerate development of speech biomarkers. 

Speech data is collected in a standardized fashion via a custom speech app, while clinical and biomarker data come from different partner clinical sites and are harmonized before they are shared with partners via the AD Data Initiative's AD Workbench. The SpeechDx dataset may inform new diagnostics, with biomarkers detecting decline as early as possible, or may provide insights into prognostics to predict future cognitive decline. 

“The SpeechDx dataset has the potential to transform the fight against Alzheimer’s by harnessing the power of speech and digital voice-based biomarkers, opening up new possibilities for earlier disease detection, and progression monitoring,” said Niranjan Bose, Managing Director, Health and Life Sciences at Gates Ventures. “This partnership with Siemens underscores our shared commitment to accelerating Alzheimer's screening and diagnostics, and we are excited for the progress we can achieve together in the coming years.” 

SpeechDx’s dataset includes approximately 2,000 participants recruited from eight clinical sites, with speech collected in three languages (English, Spanish, and Catalan). Each study participant is deeply characterized, including by medical history and demographic data and by longitudinal Alzheimer’s blood-based biomarker and neuropsychological testing and MRI imaging. SpeechDx collects data from each participant for three years to track potential cognitive decline. While SpeechDx participants range from participants with normal cognition to participants with mild Alzheimer’s, most participants are cognitively normal or at earliest stages of decline. SpeechDx is on track to complete data collection in 2028 and will release data to partners in batches, starting in 2025.

“Validation of speech as a digital biomarker has the potential to transform how we diagnose and predict Alzheimer’s disease,” said Rhoda Au, PhD, MBA, Professor of Anatomy and Neurobiology at the Boston University Chobanian and Avedisian School of Medicine, Principal Investigator of the SpeechDx clinical site at Boston University, and recipient of the 2023 Melvin R. Goodes Prize for Excellence in Alzheimer’s Drug Development. “Speech recordings are easily attainable with the widespread and low-cost use of microphones from everyday sources like cell phones, computers, and other electronic devices – leading to greater accessibility to screenings for Alzheimer’s. Combining cognitive assessments, patient demographic information, and speech data may provide a more comprehensive diagnostic and prognostic method of screening for Alzheimer’s.”