Tokyo's Sumida Ward pilots AI voice app to detect early dementia signs
In short
Sumida Ward in Tokyo has begun a pilot using an AI smartphone app that analyses a user's voice for signs of cognitive decline, which Yomiuri Shimbun reported is the first such attempt by a Japanese municipality. The app, called Hanashitene, was developed by AI startup IGSA and is offered free to residents, with 259 people registered as of the end of last month. The aim is to catch mild cognitive impairment early, before it develops into dementia.
Why it matters
The pilot tests whether a low-effort, conversational screening tool can reach older residents who may avoid conventional cognitive tests, which require correct answers. Because mild cognitive impairment can stabilise or improve, early detection could feed into lifestyle-based prevention, though the ward is still verifying the app's effectiveness. The documents present it as a local-government first in Japan, so its results may shape how other municipalities consider similar tools.
Key facts
- Sumida Ward in Tokyo has started a pilot using the AI smartphone app Hanashitene to detect signs of cognitive decline from voice, described as the first such municipal effort in Japan. [ 1 , 2 , 3 ]
- The app was developed by IGSA, an AI startup founded by people from the University of Tokyo, and was introduced as a pilot in September last year. [ 1 , 2 , 3 ]
- It offers a voice diary, a measurement function carried out every three months, and voice action training. [ 1 , 2 , 3 ]
- The measurement function analyses speaking speed, rhythm, vocabulary and grammar structure and rates brain functions such as language, memory and concentration on a four-level scale. [ 1 , 2 , 3 ]
- The service is provided free to ward residents, and 259 people had registered as of the end of last month. [ 1 , 2 , 3 ]
- According to Japan's National Center for Geriatrics and Gerontology, 5 to 15 percent of people diagnosed with mild cognitive impairment progress to dementia each year, while 16 to 41 percent return to a normal state. [ 1 , 2 , 3 ]
- The ward says the conversational format reduces the psychological resistance older users may feel toward conventional tests. [ 1 , 2 , 3 ]
Confirmed by several sources
- Sumida Ward began a pilot using the AI app Hanashitene to detect cognitive decline from voice, with 259 residents registered as of the end of last month. [ 1 , 2 , 3 ]
- The app was developed by AI startup IGSA, founded by University of Tokyo graduates, and offers voice diary, measurement and voice action functions. [ 1 , 2 , 3 ]
- The measurement function runs every three months and evaluates brain function on a four-level scale after analysing speech speed, rhythm, vocabulary and grammar. [ 1 , 2 , 3 ]
- The National Center for Geriatrics and Gerontology figures state that 5 to 15 percent of people with mild cognitive impairment progress to dementia annually and 16 to 41 percent return to normal. [ 1 , 2 , 3 ]
Still unclear
- Whether Sumida Ward's pilot is genuinely the first by a Japanese municipality All three documents attribute the first-in-Japan claim to Yomiuri Shimbun's reporting rather than to the ward itself, and no independent confirmation appears in the documents.
- Whether the app actually detects cognitive decline or lowers dementia risk The documents say the ward is still verifying effectiveness and report no results; the MCI progression figures are general statistics, not outcomes of this pilot.
- How long the pilot will run, what it costs, and whether it will be expanded None of the documents state a schedule, budget or expansion plan.