AI tool for detecting Alzheimer’s and related dementias shows promising results
A new study led by Columbia Nursing Assistant Professor Maryam Zolnoori, PhD, and published in the July 17, 2023, online issue of Artificial Intelligence in Medicine, analyzes the performance of ADscreen, a speech processing computer algorithm she is developing to support clinicians in detecting and monitoring the progression of Alzheimer's disease and related dementias (ADRDs).
Using a dataset featuring voice recordings from ADRDs patients, Zolnoori and her colleagues programmed ADscreen to look for changes in the acoustic and linguistic parts of speech. For acoustic parts of speech, the research team modeled speech fluency, frequency and spectral parameters, voice intensity, rhythmic structure of the voice, and voice quality. For linguistic parts of speech, they modeled semantic disfluency, lexical richness, syntactic structure, and psycholinguistic structure in patients’ voices. These speech parameters are the earliest signs of cognitive impairment caused by ADRDs. ADscreen flags alterations in these areas and provides extensive data visualizations on the extent of each change.
In the United States, around five million people (about twice the population of Mississippi) and 11% of older adults are affected by ADRDs. Half of patients with ADRDs remain undiagnosed and undertreated, burdening patients, caregivers, and the health care system.
“This technology is cost-effective, non-invasive, and remarkably sensitive. We found it can be integrated into clinical workflows, enabling the early detection of cognitive impairment in patients. This identification facilitates the delivery of appropriate and timely care. It's also worth noting that no commercially available algorithm utilizes audio-recorded verbal communication for cognitive impairment detection.” says Zolnoori.
Joining her in conducting the study were Associate Professor Maxim Topaz, PhD, of Columbia Nursing, and Ali Zolnour, of the University of Tehran’s School of Electrical and Computer Engineering.
VNS Health, one of the nation’s largest home- and community-based health care nonprofits, where both Topaz and Zolnoori conduct clinical research, will play a key role throughout the study in testing the practical applications of this novel technology.
Given the diverse populations served through VNS Health’s NYC-based home health services, the organization is now in the initial stages of enlisting home care nurses across the organization to further develop the technology with consenting patients in the field.
“The next phase will be to test and finetune this tool in a clinical setting with patients,” explains Topaz.
The research team was honored with two awards, totaling $1.2 million, from the National Institute on Aging. These awards will facilitate the implementation of this innovative screening algorithm at VNS Health, aimed at heightening nurses' awareness of patients' cognitive status. This early detection system will aid in the timely introduction of effective interventions, minimizing the risk of adverse outcomes related to cognitive impairment.
Research around this topic first began in 2020 when Zolnoori observed her 55-year-old aunt’s battle with Alzheimer's’ disease.
“Over a 10-year span, my aunt’s condition deteriorated despite checkups. When she was finally diagnosed, it was too late for any treatment. Early detection could have made a difference by slowing down or reversing the progression of the disease with appropriate interventions.”