THE INSTITUTEDementia affects millions of people worldwide. There is no treatment, but an early diagnosis can help patients slow the progress of their symptoms. The condition can affect people’s mental function, behavior, and memory.
Because dementia can cause different patterns of damage to the brain, no single test can determine whether someone has it. Instead, doctors use several screening tools including in-person interviews, questionnaires about daily routines, and drawing assessments. The tests, performed by clinicians and other professionals, are done regularly to check for changes—which can become expensive.
IEEE Fellow Helen Meng, a professor of systems engineering and engineering management at the Chinese University of Hong Kong (CUHK), is working on a machine-learning platform to helpmake screening more accessible and less expensive. The platform likely will use data analytics, human-computer interaction, and spoken-language technology.
Hong Kong has a large aged population, Meng says, and dementia is on the rise. Although all the region’s citizens are covered by the public health care system, it can take a long time to get an appointment with a specialist, she says, so valuable time can be lost. She is working with other researchers at the university, including many IEEE members, to make assessments accessible through AI, and eventually give people the ability to do self-assessments.
“As a researcher, a lot of our efforts have been focused on advances in existing applications such as high-accuracy speech recognition,” Meng says, “but I want to look into using the technology for new applications such as detecting early signs of dementia. The way to catch dementia early is to do frequent assessments on an individual’s capabilities. If dementia can be detected earlier, intervention can be started sooner.”
One well-known exam that neurologists perform is the Montreal Cognitive Assessment. Designed to evaluate short-term memory, language ability, and attention span, it includes activities such as naming animals and drawing components of a clock. As with other such assessments, the Montreal test is still done on paper, and the results are not digitized.
The neurologist interviews patients and asks them to assess their memory and cognitive functions. The patients’ responses might be subjective, varying from day to day even if their abilities don’t.
Meng says machine learning and big data can help make those diagnoses more objective. Artificial intelligence algorithms and other technology could automatically analyze collected data.
In particular, spoken-language technology could be used to assess a person’s cognitive health and emotional state based on their speech.
“We want to be able to identify spoken-language biomarkers that are indicative of neurocognitive disorders,” Meng says. “The reaction time after a question is asked could be recorded. For example, if there’s a lot of hesitation or pausing, even at millisecond intervals, these could be measured in an objective way using engineering approaches.”
Offering tests on a computer or recording people’s speech while they answer questions via a telephone could help reduce the number of needed visits to the doctor, Meng says. A clinical cognitive expert or neurologist would review the automated assessments.
“We don’t intend for our automated software to make decisions about whether someone has dementia,” she says. “Our objective is not to replace the clinicians. We look at AI as a decision-support tool.”
The project has recently been awarded the theme-based research scheme of Hong Kong’s Research Grants Council. This is among the highest level of research funding in the region, according to Meng.
COMBINING TWO PASSIONS
Meng, who grew up in Hong Kong, was accepted to medical school as well as MIT’s engineering program. She says she thought it would be a good experience to study abroad, so she attended MIT, where she earned bachelor’s and master’s degrees in electrical engineering. She also got a Ph.D. in electrical engineering and computer science there.
She joined the CUHK in 1998 and established its Human-Computer Communications Laboratory the following year. She founded the university’s Stanley Ho Big Data Decision Analytics Research Center in 2013 and serves as one of its directors.
She has collaborated on several biomedical engineering research projects with the doctors at Prince of Wales, the CUHK teaching hospital.
Meng joined IEEE in 1998, when she was an assistant professor. She served as reviewer and then associate editor for the IEEE Signal Processing Society’s Transactions on Audio, Speech, and Language Processing and eventually was elected editor-in-chief of the publication, serving in that capacity from 2009 to 2011. She was a member of the society’s board of governors and its nominations and appointments committee.
“IEEE is a global platform, so there are many ways to participate,” she says. “I’ve made quite a few friends and met colleagues around the world who are experts in their area. It has been a great experience.
“Membership also broadens one’s horizons. Through IEEE’s conferences and publications, you get to look beyond your own area of expertise.”
Meng works to increase the number of women in engineering. She and other women have spoken during the annual IEEE Signal Processing Society conference’s luncheon.
“We make sure that female keynote speakers are invited to conferences, not just men,” she says. “We need more gender diversity.”