how to protect the memory of alzheimer's patients? microsoft's "remember me" makes cognitive training more personalized and more daily
2024-09-23
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
picture description: the interface of the "remember me" app. photo provided by the interviewee
september 21 is world alzheimer's day. with the aging of the global population, dementia represented by alzheimer's disease is considered one of the biggest health crises in the 21st century. although there is no drug that can completely cure the disease clinically, early diagnosis and early intervention can help delay the decline of cognitive function, and cognitive training has also been proven to be an effective means to alleviate the progression of the disease.
how can we protect the memories of the "old children" trapped in the maze of time? last weekend, at the "protect cognition, act now world alzheimer's day public science popularization event" held by shanghai mental health center, microsoft research asia and xujiahui academy, the reporter learned that microsoft research asia is currently working with shanghai mental health center to carry out joint research. with the help of multimodal large models and intelligent agent (ai agent) technology in microsoft azure openai services, a personalized cognitive training framework "reme" has been developed, which expands the training scope of automated cognitive training and provides a new method for digital cognitive training, which is expected to help delay cognitive decline.
existing cognitive training is relatively simple
"cognitive impairment affects various types of memory, including working memory, episodic memory, and semantic memory. however, the design of existing training tasks is relatively simple, which limits the breadth of training content. many elderly people have also reported that the game content is "boring" and "uninteresting." yue ling, deputy chief physician of the department of geriatrics at shanghai mental health center, said that for patients with cognitive impairment, persistent training is crucial to alleviating symptoms. once the training becomes "boring" and "uninteresting", it is easy for people to give up.
the reporter learned that the current clinical cognitive training mainly relies on caregivers or therapists to carry out themselves, or use software-driven digital therapy. the former relies on a lot of manpower, is costly, and imposes a heavy burden on caregivers; the latter is often similar to standardized tests, lacks flexibility, and patients can only answer preset puzzles, which is far from the memory training required to improve cognition. in addition, the existing digital therapies are not convenient, interactive, and intuitive in software application, which is also affecting patient compliance (the patient's behavior of following the doctor's prescription for treatment). "software-driven cognitive training mainly focuses on working memory training, while episodic memory and semantic memory training, which are closely related to the patient's daily life, are rarely involved. especially for patients with alzheimer's disease, the first memories to be damaged are often those related to the self." yue ling said.
the multimodal big models in the field of artificial intelligence are constantly developing and showing many new "capabilities", such as providing instant multimodal analysis, realizing open questions and answers based on world knowledge, providing emotional voice interaction, integrating and processing multiple sensory data, etc. "dementia, which is what we often call cognitive impairment, lacks cognitive abilities in thinking, memory, attention, understanding, judgment and language, right? why don't we give the 'all-rounder' multimodal big models a try?" therefore, researchers at microsoft research asia proposed this bold idea and began to discuss and cooperate with the department of geriatric psychiatry of shanghai mental health center.
personalized training close to daily life
"what did you have for breakfast today?" "what typhoon has hit shanghai recently?" uncle zhang experienced "reme" at a popular science event. through mobile phones or wearable devices, users can upload personal memory content, and then use human-computer interaction to conduct individualized episodic memory or open memory training of world knowledge. yue ling said: "'reme' can respond interactively in real time, support input and output in multiple modes such as text, images, and voice, and provide patients with cognitive training experience in the form of a conversational robot. its training is closer to daily life, helping patients train in many aspects, such as working memory that processes temporary information, episodic memory involving self-experiences, and semantic memory covering factual knowledge."
in addition, researchers have tried to use wearable devices based on sensor technology to record important nodes and events in daily life, making personalized memory training more convenient, and providing a training framework so that doctors and other researchers in this field can use built-in cognitive game templates to quickly create personalized training games. "remember me" will also provide an evaluation report based on each interactive activity, so that doctors or rehabilitation therapists can track the patient's cognitive level and adjust the training tasks in a timely manner as needed.
"it is a challenge for us to combine the general capabilities of large multimodal models with professional knowledge and apply them to the field of cognitive training to create a professional domain intelligent agent," said wang zilong, a senior researcher at microsoft research asia. "but through a series of previous studies, we can now enable intelligent agents to use newer and more comprehensive domain knowledge, maintain their cutting-edge capabilities in specific fields, and better interact and train tasks."
the push for more clinical evidence
although ai-assisted cognitive training is still in the research and development stage, it does bring a glimmer of hope. "in the future, shanghai mental health center plans to design new methods of cognitive training based on this framework and conduct a series of rigorous clinical experiments to verify the effectiveness of this personalized cognitive training framework and obtain more clinical evidence," said yue ling.
dr. lili qiu, vice president of microsoft research asia, said that research and innovation in artificial intelligence technology in the medical field are helping medical institutions improve the quality and efficiency of medical services and provide patients with more accurate and personalized treatment plans. she also revealed that as clinical evidence gradually accumulates, microsoft research asia will continue to improve and iterate the personalized cognitive training framework "remember me", hoping that the tool will gradually expand its scope of application and eventually help improve the health of patients with cognitive impairment in community and family environments, reducing the burden on caregivers and family members. "in addition to cognitive training, agent intervention technology based on multimodal large models is expected to be applied to a wider range of fields in the future, such as emotion regulation, habit improvement, and auxiliary interventions for autism."
xinmin evening news reporter ma dan
