With the rising number of dementia cases worldwide, it is becoming more essential than ever that new cases be detected as early as possible. Still, given the difficulty of distinguishing early signs of dementia from the natural forgetting that all seniors experience, formal diagnosis can often be cumbersome due to the need for conventional medical imaging which can be time-consuming and expensive.
But a new artificial intelligence test developed by researchers at Osaka University may provide a low-cost solution that can detect dementia far earlier than once thought possible. Using machine learning, the test features a series of questions asked by interactive computer avatars to detect speech and language patterns found in dementia patients.
Standard questions administered can include: "What's your name," "What's the date today," and "How is your appetite." Along with being intended to put patients at their ease, these questions also allows the computer to measure gaps in conversation and other speech and facial patterns. The system uses a computer avatar resembling a human-like female character speaking in a pleasant voice. Designed to be interactive, the avatar nods its head whenever the computer recognizes patient input used. The system also includes a small moving dot on which patients are directed to gaze as a measure of reaction time. Along with standard questions, the system also administers brief memory tasks. From the information presented by patients, the machine algorithm combines data on speech patterns, eye gaze, and percentage of nouns and verbs used.
What the researchers found was that the computer could detect dementia cases 90 percent of the time using just six questions (at two to three minutes per question). The new system is also designed to vary the questions asked and to give new questions to people returning for later visits to avoid second-guessing. While this research is still in its early stages, lead researcher Takashi Kudo emphasized the potential value of this test. "If this technology is further developed, it will become possible to know whether or not an elderly individual is in the early stages of dementia through conversations with computer avatars at home on a daily basis.” he said in a recent interview.
Since this system was originally developed for Japanese patients adapting it to be used in other languages will require further research. Also, given that facial expressions are often culture-dependent, researchers may also need to develop new algorithms to take these differences into account. Still, while more research needs to be done to validate this approach, the use of computer-driven diagnostic systems represents an intriguing new tool for identifying dementia patients in the early stages. Given the high number of dementia cases expected over the next few years, tools like this will be badly needed.