Li Feifei team developed a home AI system to monitor the symptoms of new crowns for elderly people living alone

Caring for the elderly became more difficult during the new coronary pneumonia pandemic. Does artificial intelligence play a role in this area? On April 6th, local time, in a live broadcast from Stanford University’s Institute of People-oriented Artificial Intelligence (HAI), Li Feifei, a professor of computer science at Stanford University, introduced the Artificial Intelligence home system, which tracks the health of residents, including symptoms of new coronary pneumonia, while also ensuring privacy.

(Original title: Li Feifei team is developing a home AI system to monitor the symptoms of the new crown of the elderly living alone)

Journalist Wang Xinxin

Li Feifei team developed a home AI system to monitor the symptoms of new crowns for elderly people living alone

Li Feifei

The aI system is designed to help older people, most of whom live alone, stay in touch with their families or health care providers. The best way to protect older people is to reduce contact with people, especially those with new coronary pneumonia who have not shown symptoms. The advantage of the home system, according to Li Feifei’s team, is that it allows caregivers to remotely monitor the elderly’s existing illness and basic health conditions, reducing the risk of exposure.

In a live speech, Ms. Li and her team said the system was being developed by an interdisciplinary team of clinicians and computer scientists before the outbreak of new coronary pneumonia. “Over the past few years, we’ve been working on an AI system that helps older people live independently and manage their chronic diseases. Recently, we realized that this technology could also help older people in the new coronary pneumonia epidemic. Li Feifei said in his speech.

According to Li Feifei, the entire home AI system includes cameras and smart sensors installed in the home. In his speech, Li mentioned four sensors, including a camera, a depth sensor, a thermal sensor and a wearable sensor. The team’s research focused on the first three. Because privacy is so important in this system, research for cameras is even more challenging. “The camera can reveal details of personal activity, but it doesn’t match most people’s privacy needs. Li Feifei said.

How does the entire system work and how does privacy ensure? Li Feifei introduced one by one in his speech. When the sensor obtains the data, it is sent to a secure central server for processing. In the process, however, Li feifei also acknowledged that there are security risks at this stage, such as the threat of cyberattacks. But she stressed that researchers follow privacy and security guidelines throughout the process. The team equipped edge devices with encrypted disks that remove data that involves user privacy, obfuscated faces, and encrypted them before being transferred to the cloud.

Once the data reaches the server, a team of clinicians and AI experts analyzes and annotates it to develop machine learning models. The trained model identifies some clinically relevant behaviors, including breathing, sleep, diet, and other behaviors. Mr Li said the team was currently developing a model involving daily activities that could calculate whether the user’s health had deteriorated. But this model is not an in-depth and extensive analysis of all the daily activities of the user, and needs to find a balance between privacy and public safety.

The trained model can be deployed to edge devices and run locally. In this way, the research team set up a closed-loop system, data security can also be guaranteed. However, this closed-loop system cannot further update and upgrade the model. To address this, Li feifei said, the team is envisioning a way to use joint learning and unsupervised learning, where models on each edge device are updated without manual annotation to use the new environment and improve robustness. Through joint learning, teams can limit security attacks to devices to reduce privacy and security threats to the cloud.

Finally, the system needs a way to pass the results of intelligent sensor tests to health care workers or family members. Mr. Li said the team has not yet found a specific solution, but is considering using a mobile app or web interface.

“These sensors are not intended to make diagnostic decisions or replace clinicians, but rather to be continuous, keepaning around our elderly at home and alert clinicians and families in a timely manner.” At the end of his speech, Mr. Li said, “Of course, we have to take a holistic approach to ethics, privacy and security at every step of this study and in the deployment of this technology.” “

The challenges posed by the current pandemic of new coronary pneumonia include not only ensuring the safety and health of older persons, but also broader and urgent tracking of diseases and populations that should be isolated. Asked if the system could solve the problem, Mr Li said the team was reluctant to get involved. “Our goal is to propose cutting-edge computer vision and machine learning technologies to help address some of the most important and challenging issues in healthcare, as well as security guidelines for ethics, privacy, and AI healthcare research. Li Feifei said.

At present, the project is still in the research stage. The entire team also needed to complete the data set construction and modeling work, and the team did not say how much time it would take to complete it. However, the team has worked with On Lok, a U.S.-based premium care company, to complete a pilot study at an assisted living facility in San Francisco and will move on to the next phase.