Vocal AI can sub for professionals to give and grade elderly cognitive exams

When paired with willing older adults, talking AI can administer and assess preliminary dementia tests in much the same way—and with similar effectiveness—as human neuropsychological specialists.

The technique proved its utility in a small study conducted at the University of Lethbridge in Alberta and published online May 17 in Acta Psychologica. 

Neuroscience PhD candidate Daniela Aguilar Ramirez and colleagues had such a device give three separate tests to 12 women and 12 men between the ages of 63 and 78. The tests consisted of memory tasks involving logical memory, poetry recall and verbal working memory/attentiveness (i.e., “digit span”).  

The team found the device—a smart speaker called ABRA, for Audio Board for Robotics and Automation—accurately identified age-related memory changes. In their study report, they remark that using the technology in real-world settings might help relieve workloads for geriatric specialists, broaden access to such care for underserved populations and facilitate ongoing screening for neurodegenerative conditions. 

In their discussion section, Aguilar Ramirez and co-authors share several key observations from the project. Excerpts:  

1. Vocal AI can identify age-related changes even with a relatively small sample. 

Also notable is the finding that, when collected by ABRA, the scores of logical memory tests—both immediate and delayed—agree with those administered by humans, the authors report.

‘These findings support our hypothesis that an AI-powered speech device can perform [neuropsychological test batteries] much like a human examiner.’

2. Automated speech recognition for elderly cognitive assessment offers numerous benefits.  

The advantages go beyond mere efficiency gains for healthcare professionals, the authors remark. 

“First, assessment via AI offers a natural and comparable way to an in-person test administrator,” they write. “Second, use of a speech device breaches any unfamiliarity with technological devices the older adult patient or participant may have. Third, and most importantly, remote administration enables inclusion of hard-to-reach populations and facilitates the continuous screening of older adults.” More: 

‘This, in turn, could help early identification of prodromal stages of neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease.’ 

3. Vocal AI’s dynamic, real-time response capabilities open the door to implementing tasks that are more ecologically valid.

“While this study was investigatory in nature, we were pleased to find that the barriers of use were lower than we expected: The older population was capable of understanding and using the technology,” Aguilar Ramirez and colleagues note. “Future studies should focus on examining the reliability of other neuropsychological tests administered through this device or similar speech devices.” 

‘Harnessing the potential of this technology can assist in addressing present and future healthcare challenges more effectively.’

4. Despite its promise, the use of vocal AI for assessing cognitive status in older adults presents several challenges that warrant careful consideration. 

“A central issue stems from the current generation of older adults’ limited knowledge and understanding of AI and its mechanisms,” the authors point out, citing prior research showing a significant preference by participants for human examiners over AI devices. 

‘With respect to our study, some participants expressed similar concerns, although others indicated that they would use the device for cognitive games, medication tracking and daily health checks.’ 

5. Encouragingly, there is evidence that older adults entertain a willingness to learn about AI and its practical applications. 

“It is crucial to develop targeted educational programs that demystify AI, specifically explaining how devices such as ABRA function and contribute to assessments may reduce negative perceptions and promote greater acceptance,” Aguilar Ramirez and co-authors write. 

‘Fostering this understanding will be critical to ensuring the successful adoption of AI-based tools in neuropsychological practice.’

The study is posted in full for free. 
 

• Recent research in the news:
   
  • Chapman University: AI-powered app enables anemia screening using fingernail selfies 
     
  • West Virginia University: Researchers test AI’s limits in emergency room diagnoses
     
  • Keio University: AI tool enables real-world assessment of eczema severity via smartphone photos
     
  • Harvard/Mass General Brigham: Machine learning model helps identify patients at risk of postpartum depression 
     
• Funding news of note:
   
  • AI-powered drug development startup Pathos AI closes $365M investment 
     
  • PhaseV lands $50M Series A to supercharge AI for clinical development
     
  • AI biotech firm SpotitEarly raises $20.3M to scale noninvasive cancer detection in US
     
  • Intrepid Labs raises $7M to expand AI-driven drug formulation platform
     
  • Indian healthcare AI startup Qure.AI aiming for IPO in two years, CEO says
     
• From AIin.Healthcare’s news partners:
   
  • Cardiovascular Business: AI delivers real-time blood clot monitoring, helping cardiologists make treatment decisions
     
  • Health Imaging: Newly launched algorithm reduces image artifacts in real time during IR procedures