Verified by 
Neurological Assessment
AI Predicts Dementia Risk from Speech Years in Advance
Alzheimer’s disease risk predicted from voice recordings non-invasively by AI.
Posted July 4, 2024 Reviewed by Kaja Perina
One of the most promising uses of the predictive powers of artificial intelligence (AI) machine learning is to provide clinicians with early disease detection to improve patient care. Last week researchers at Boston University published a study in Alzheimer’s & Dementia, the journal of the Alzheimer’s Association that unveils an AI capable of predicting the risk of a patient with mild cognitive impairment developing dementia within six years non-invasively from speech analysis.
“Identification of individuals with mild cognitive impairment (MCI) who are at risk of developing Alzheimer's disease (AD) is crucial for early intervention and selection of clinical trials,” wrote corresponding author Ioannis Paschalidis, PhD, a professor of electrical and computer engineering and director of computing and computational science and engineering, along with co-authors Rhoda Au, PhD, a professor of medicine, anatomy and neurobiology, Vijaya B. Kolachalama, PhD, FAHA, an associate professor of medicine and assistant professor of computing and data science, Cody Karjadi, Jingmei Yang, Boran Hao, and Samad Amini at Boston University.
People are living longer, and age is the greatest risk factor for dementia as those aged 65 and older comprise a majority of cases. Globally will be 1.5 billion people aged 65 or older by 2050 according to the 2022 Revision of World Population Prospects report by the United Nations Department of Economic and Social Affairs Population Division. By 2030, an estimated 73 million, or every 1 in 5 Americans, will be age 65 and over according to a report by the U.S. Federal Interagency Forum on Aging-Related Statistics.
Dementia, also known as major neurocognitive disorder, is a chronic condition that may impair cognition, memory, behavior, and emotion. It is an umbrella term for a group of symptoms and the impairment of cognitive abilities, such as memory, communication abilities, focus, reasoning, judgement, attention, decision making, problem solving, and visual perception, that is severe enough to interfere with daily life activities.
In 2013, the term dementia was absorbed under major neurocognitive disorder in the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM-5) by the American Psychiatric Association (APA). Major neurocognitive disorder may be caused by Alzheimer’s disease (AD), traumatic brain injury, Parkinson’s disease (PD), Huntington’s disease, Lewy body disease, frontotemporal degeneration (FTD), cerebrovascular disease, hippocampal sclerosis (HS), mixed pathologies, substance or medication use, prion disease, and more.
The most common cause of major neurocognitive disorder is Alzheimer’s disease, which accounts for roughly 60%-80% cases according to the 2024 Alzheimer’s Disease Facts and Figures report by the Alzheimer’s Association. The National Institute of Health’s (NIH) National Institute of Aging estimates there are over six million Americans with Alzheimer’s, and a majority are 65 years or older. An estimated two-thirds of all U.S. Alzheimer’s cases are women, and every 65 seconds, an American develops Alzheimer’s disease according to data from the Women’s Alzheimer’s Movement and Cleveland Clinic.
The need for diagnostic tools for early intervention is a growing worldwide need. Worldwide, the number of people with dementia is expected to reach 152.8 million cases by 2050 according to research published in 2022 in The Lancet Public Health. An estimated four million Americans will have dementia by 2060 according to the U.S. Centers for Disease Control and Prevention (CDC).
The researchers point out numerous advantages to using AI to analyze voice recordings as a diagnostic. It is a method that is non-invasive, cost-effective, widely accessible, easily adaptable for remote assessments, and easy-to-perform.
“Speech during cognitive exams has been identified as a promising biomarker that strongly correlates with underlying cognitive dysfunction,” the researchers wrote.
The team used data from the Framingham Heart Study, a longitudinal, population-based, multi-generational cohort study of chronic disease that began in 1948 by the U.S. Public Health Service with participants in Framingham, Massachusetts. In 2005, FHS digitally recorded in-person interviews of neuropsychological tests (NPT) that include the Wechsler Memory Scale (WMS), Boston Naming Test (BNT), and Hooper Visual Organization Test.
Voice recordings from 166 study participants (107 females, 59 males) ranging in age from 63 to 97 years old from the Framingham Heart Study (FHS) contained roughly one-hour interviews of neuropsychological tests that assess attention, memory, naming, language, visuoperceptual skills, abstract reasoning. Health risk factors, education, and apolipoprotein E (APOE) gene alleles. For certain populations, the APOE-e4 gene increases the risk of developing Alzheimer's disease. From the study participants, 90 progressed to AD dementia and 76 remained stable, with minor cognitive impairment, within a six-year period.
Rather than use acoustic analysis, the team opted to use text-based analysis. The scientists combined AI machine learning with natural language processing to create a fully automated pipeline for converting raw voice recordings into meaningful predictions.
“Our best models, which used features generated from speech data, as well as age, sex, and education level, achieved an accuracy of 78.5% and a sensitivity of 81.1% to predict MCI-to-AD progression within 6 years,” reported the scientists.
This proof-of-concept may one day enable clinicians to provide early intervention and identify suitable candidates for clinical trials for developing novel Alzheimer's disease pharmaceutical drugs. The application of AI models to health care diagnostics is an area that can help improve patient outcomes in the future.
Copyright © 2024 Cami Rosso
