Yunqi Information: [ Click to view more industry information ]
Here you can find the first-hand Shangyun information in different industries, what are you waiting for, come soon!
The elderly are high-risk groups in the epidemic. According to a report released by the US Centers for Disease Control and Prevention on March 18, in the United States, about 80% of the deaths of new coronary pneumonia are older than 65 years. During the isolation period, it is particularly important to monitor the health of the elderly in real time.
On April 1, at the live broadcast of "New Coronary Pneumonia and AI" hosted by the Stanford University People-oriented AI Research Institute (HAI), Dr. Li Feifei, a professor at Stanford University, former vice president of Google, and HAI co-director, proposed the concept of a home AI system The system can track the health status of residents on the basis of ensuring the privacy of residents, and monitor the symptoms of new coronary disease.
Before the online event began, Li Feifei published a message in the circle of friends, saying, "Science knows no borders, and sickness knows no borders. Many Stanford University researchers are involved in the research of new coronary pneumonia. Only AI-related diseases include disease Diagnosis, treatment, epidemic prevention, public health, government policies, laws and regulations, and even studies on human influence, regional discrimination, social justice, press freedom, etc. This also includes the AI and medical treatment of my own Stanford laboratory in the past decade Health research (especially hand hygiene and home disease management).
The purpose of the design of the home AI system is to benefit the elderly, especially the elderly living alone, so that they can better enjoy the care of family members or medical workers. During the outbreak of the new crown epidemic, the best way to protect the elderly is to reduce contact with other people, even those who care for their pneumonia symptoms at all times. The system is designed to track the health status of the elderly in real time while reducing the risk of external contact. At the same time, the nursing staff can remotely monitor the basic physical condition of the elderly because many elderly people have various health problems.
In the report, Li Feifei and his team stated that their interdisciplinary team was composed of clinicians and computer scientists, and the project was already underway before the outbreak of the new crown. "In the past few years, we have been studying how AI technology can help older people live more independently and better cope with chronic diseases. But recently, we have realized that AI technology, which is also used for long-term care, is responding to acute epidemics like the New Crown It also helps the elderly. "She said.
The project is still in the research stage. The team needs to complete the data set construction and model training, and it is unclear how long it will take to complete. However, the system was originally designed for elderly care, and it will be an ideal choice for health monitoring in the era of large-scale isolation.
Privacy and security issues of home AI systems
The system will consist of cameras and smart sensors installed at home. The report outlines four sensors-camera, depth sensor, thermal sensor and wearable sensor (eg FitBit). Li Feifei said that their research is currently focused on the first three areas. She admits that privacy protection is essential in such a system, so the setting of the camera brings greater challenges. "Camera sensors record a lot of detailed information about individual activities, so they are most likely to touch people's privacy needs," she said.
When the sensor captures the data, the system sends the data to a secure central server. Li Feifei admitted that there are inherent security risks in this process, such as cyber attacks. In an email replying to VentureBeat, she emphasized that the researchers followed privacy and security guidelines throughout the process. "For example, our edge devices are equipped with disk encryption, data will be deleted for privacy attributes (such as facial obfuscation), data will be encrypted before transmission to the cloud, and our servers comply with the Health Insurance Portability and Accountability Act (HIPPA), "she says.
Once the data reaches the server, clinicians and artificial intelligence experts in the team will analyze and annotate it to develop a machine learning model.
"Then we trained AI models to identify various clinically relevant behavioral patterns, including breathing, sleep, diet and other behaviors," Li Feifei said. The behavior patterns they focus on are those that may cause deterioration of health in daily life activities. In other words, the focus of this model is to find specific metrics. This is not an in-depth and extensive analysis of all daily activities of residents. Li Feifei said that the significance of training AI models is to achieve a balance between practicality and privacy.
The team then deployed the trained model to the edge device, where the monitoring system can be run locally. This will create a closed loop system so that no data can be leaked out. This essentially ensures data security, but prevents any further training on this model.
The researchers came up with a solution to this limitation, and Li Feifei outlined it in an email to VentureBeat. "We envisage that the model on each edge device will continue to be updated to adapt to the new environment, and through federated learning, to improve robustness in an unsupervised manner. Through federated learning, we limit the targets of cyber attacks to devices Itself, not the device and the cloud, to reduce privacy and security risks. "
Federated Learning is an emerging basic artificial intelligence technology. It was first proposed by Google in 2016. It was originally used to solve the problem of Android mobile phone end users updating models locally. Its design goal is to ensure the exchange of big data. Under the premise of protecting information security, protecting the privacy of terminal data and personal data, and ensuring legal compliance, high-efficiency machine learning is carried out between multiple participants or multiple computing nodes.
In the final step, the system also needs to pass the detection results of the smart sensors to medical staff or family members of the elderly. Li Feifei said that the team has not found a specific solution, but is considering using APP or web interface, both of which can ensure data security through dual authentication.
She emphasized: "These sensors are not used to make diagnostic decisions or replace clinicians, but they can provide real-time and continuous monitoring of the elderly living in the home and promptly issue health warnings to clinicians and family members."
"Of course, at every step of this research, and during the deployment of this technology, we must think deeply about each of these ethical, privacy, and security issues," she added.
The current outbreak should not only focus on the safety and health of the elderly, but also pay close attention to the situation of other patients and isolation personnel. Some components of the system can be adjusted to track without violating citizens ’rights and privacy. But Li Feifei is temporarily unwilling to get involved in these fields, she believes, "Our goal is to use the most cutting-edge computer vision and machine learning technology to help solve some of the most important and challenging health care issues, and for artificial intelligence health care The study proposes a guide to ethics, privacy and security. "
Li Feifei said that the current research has progressed to the next stage. They have completed a pilot in a nursing home in San Francisco, California, and cooperate with a local nursing facility called On Lok, which is committed to providing high-quality nursing services for the elderly.
Wearable devices enable non-contact monitoring
Some other home AI monitoring systems also involve wearable devices, such as Current Health, iRhythm, and LiveFreely. For example, the Zio ECG patch developed by iRhythm can be worn continuously for 14 days and can provide continuous ECG monitoring. The Care.ai system uses computer vision technology to achieve non-contact monitoring. The concept is similar to the Li Feifei team, but the Care.ai system is mainly for hospitals, not home care.
From another perspective, the current social segregation policy makes old people living alone also more lonely. In addition to technical monitoring, for the elderly, the more important thing is family care and (online) companionship!
Reference link:
https://venturebeat.com/2020/04/06/stanford-researchers-propose-ai-in-home-system-that-can-monitor-for-coronavirus-symptoms/
https: //hai.stanford .edu / events / covid-19-and-ai-virtual-conference / overview
https://www.zhihu.com/topic/20935178/intro
[Yunqi Online Classroom] Every day, product technology experts share!
Course address: https://yqh.aliyun.com/liveJoin the community now, face-to-face with experts, and keep abreast of the latest developments in the course!
【Yunqi Online Classroom Community】https://c.tb.cn/F3.Z8gvnK
Original link
This article is the original content of the Yunqi community and may not be reproduced without permission.
