Imagine a world where a recording of a person’s voice could help healthcare professionals more quickly detect a neurological condition.

Thanks to efforts across the Clinical Innovation, Data Science and Neuroscience teams at Janssen, along with other internal and external strategic partners, this reality may now be close at hand. Recent developments in speech biomarker technology may be able to help predict results of current diagnostic procedures for Alzheimer’s disease (AD). This discovery has the potential to improve clinical trials and patient outcomes.

In March of this year, I presented with collaborators INRIA, University of Edinburgh and ki:elements at the South By Southwest (SXSW) Conference on the implications of using speech biomarker technology for AD and other neurological diseases. Once a festival for music, film and multimedia, SXSW has evolved over its 35-year history to showcase prominent developments in healthcare and technology and muse on their future implications.

At our panel, AI is Listening—And it Says You Have Alzheimer’s, Dr. Alexandra König of INRIA noted that the brains of patients with Alzheimer’s disease are seen to show physiological changes in line with the disease even 20 years before they start to display symptoms of cognitive impairment.¹ With this staggering finding in mind, tools that help clinicians to assess people early on for evidence of the disease with less reliance on invasive procedures is welcomed.

Traditionally, AD requires clinicians to look at the brain through medical imaging such as PET scans to diagnose the disease and understand how far it has progressed. Since these procedures require patients to be on-site with trained medical technicians and tailored equipment, it is crucial to have an effective pre-screening process to determine if patients meet the requirements for an Alzheimer’s disease clinical trial.

Enter speech biomarker technology. Biomarkers are biological indicators that determine if a disease is present and/or how far it has progressed in a person. In this case, speech patterns such as changes in pitch and volume, pauses between words and phrases, and other vocal features are measured by machine learning models to identify signs and severity of AD. Janssen is currently applying this technology as one component of the screening process for certain clinical trials.

It works like this: in a trial, participants can respond to speech prompts through a 10-15-minute automated phone call at home, which may help researchers determine more quickly which participants could be eligible to enroll. This could also reduce the amount of time and cost that enrolled participants will spend throughout a trial to travel to sites. After enrollment, patients complete routine clinical assessments so researchers can understand the effects of the experimental treatment. With remote digital monitoring like a speech biomarker tool to support clinical assessments, trials may be able to be designed to have fewer site visits.

Our mission to improve all aspects of clinical trials is at the heart of Janssen Clinical Innovation’s work. Since 2018, we’ve led this collaboration with the Data Science team and various therapeutic area teams at Janssen, as well as with collaborators like ki:elements, which created the foundational speech analysis software, and with research collaborators INRIA, Maastricht University and University of Edinburgh. Thanks to this effort, Janssen is currently leveraging its Autonomy Study to further evaluate speech biomarker’s effectiveness as a pre-screening tool in trials. Autonomy itself will measure the safety and efficacy of a drug candidate for early-onset Alzheimer’s disease.

This work is a powerful reminder of the importance of collaboration across technology, data science, research and other disciplines. We aim to produce innovative, valuable tools that can potentially aid in easier, more efficient diagnosis and treatment of neurological disorders. We envision a future where this speech biomarker technology could integrate with digital therapeutics to help monitor and optimize sleep, diet, exercise, and games to improve cognitive function for patients. It’s our hope that speech biomarkers, and other machine learning technologies, can easily integrate into patient lifestyles for more automated, adaptable treatment modalities that optimize patient outcomes at every turn.

Below is a narrative I presented at SXSW as an illustration of how this speech biomarker technology could be part of a “brain health package” in the future that helps to optimize patient health.

References

1. Hampel H, Blennow K. CSF tau and β-amyloid as biomarkers for mild cognitive impairment. Dialogues Clin Neurosci. 2004;6(4):379-390. doi:10.31887/DCNS.2004.6.4/hhampel

April 13, 2022