Researchers have developed an AI-powered speech analysis tool that accurately detects early signs of dementia.
Developed by experts from UT Southwestern Medical Center, the early screening tool effectively identified mild cognitive impairment and dementia in a Spanish-speaking cohort.
Munro Cullum, PhD, the corresponding author of the research, emphasised the potential of the technology: “Analysing a sample of speech obtained during some brief, routine neuropsychological tests shows promise in our ability to quickly screen for signs of cognitive impairment, particularly in population-based research studies.
“Machine learning-based tools such as this may play an increasingly important role in the future of cognitive screening for dementia.”
What is dementia?
Dementia is a broad term encompassing various brain disorders characterised by a decline in cognitive function severe enough to interfere with daily life.
It affects memory, thinking, orientation, comprehension, calculation, learning capacity, language, and judgment.
Alzheimer’s disease is the most common cause of dementia, accounting for 60-70% of cases, followed by vascular dementia, dementia with Lewy bodies, and frontotemporal dementia.
The global impact of dementia is substantial. According to the World Health Organization (WHO), over 55 million people live with dementia worldwide, and this number is expected to rise to 78 million by 2030 and 139 million by 2050 due to ageing populations.
The societal and economic burden is profound, with an estimated cost of $1.3 trillion annually, projected to double by 2030.
Research highlights the need for better diagnostic tools, effective treatments, and supportive care systems to manage the growing dementia crisis.
Now, researchers are leveraging the untapped potential of AI to improve dementia diagnostics.
Training the AI tool
To develop the AI speech analysis technology, the team collected data from 195 Spanish speakers who were part of a clinical trial in Spain.
All participants were initially evaluated and classified as having normal cognition, mild cognitive impairment (MCI), or dementia. Due to incomplete data or poor audio quality, 21 participants were excluded.
The final cohort included 174 participants with a mean age of 74, slightly more females (56%) than males. They were split into a training group of 122 (70%) and a test group of 52 (30%).
Researchers trained independent machine learning models using data from the training group participants, who completed four language tasks.
Neuropsychological performance and audio recordings were collected via the AcceXible platform, a proprietary web-based tool for disease detection through speech analysis.
Detecting early signs of dementia
The final model of the speech analysis algorithm was then used for the test group, and it was able to differentiate cognitively normal participants from those with dementia or MCI with an overall accuracy of 88.4% and 87.5%, respectively.
The final model outperformed one of the current standard-of-care screening measures known as the Mini-Mental State Examination (MMSE).
