The Embedded Machine Intelligence Lab at ASU is integrating AI into wearable technology to both personalize and assist in the monitoring of users' health and safety.
The lab aims to create complex and compact technology people can bring and wear everywhere to improve portable wellness systems. Hassan Ghasemzadeh, the director of EMIL and a professor in the College of Health Solutions, said the lab is designing, developing and validating technologies that can be successful and widely adopted.
"The focus has been building systems that use sensors that are powered by AI in a health domain ... to improve health, to detect health events, to provide automated intervention, to extract biomarkers associated with different health outcomes," Ghasemzadeh said.
Ghasemzadeh said he has been researching digital health since 2007 and established EMIL at the University in 2021. He added the lab strives to ensure transparency and provide user-centered feedback and assistance while allowing those who use the technology to maintain autonomy and control.
He said to improve the AI systems, EMIL is focused on personalization and considering individual differences. Such factors include age, demographic and user data that changes in real time.
"(We are) making sure that you make all these predictions and health assessments while taking into account the context of the user," Ghasemzadeh said. "What happens to them today? What's the situation over time? And within that context, you're able to make the prediction."
Reza Rahimi Azghan, a doctoral student studying computer science, said he focuses on algorithm development for the lab. He works primarily on "lightweight" systems that can be integrated into small devices, such as wristbands, that do not have the resources to run larger programs.
"The reason why we want these models to be run on these devices is because we want to deploy these models or deploy these systems and machines in mobile applications," Rahimi Azghan said. "On the kind of devices that directly interact with human beings."
He said the task of the machines and algorithms is to be able to predict or identify what kind of activity the user is performing at a given time. This goal can range from collecting heartbeat data, electrodermal activity levels, skin conductivity and more.
READ MORE: AI in healthcare: Balancing innovation and consequence
A common issue of most machine learning or AI models is that they are good at generalizing but struggle to retain massive amounts of data for long periods of time, otherwise known as catastrophic forgetting. However, EMIL is interested in continual learning in AI.
Rather than teaching its models to search for new knowledge, Rahimi Azghan said EMIL emphasizes increasing the models' retention to build upon or memorize the user's data on their personal device.
One of EMIL's current projects, HeatMind, is studying heat stress and heat-related illnesses firefighters are introduced to in the high-intensity environments they regularly experience. Ashley Nakawatase, a senior studying biomedical informatics and data science, started working on the project as a part of her capstone.
The lab aims to research how to monitor firefighters' health indicators in real time through wearable sensors, Nakawatase said. This education and awareness allows first responders to serve the public more safely and efficiently, she said.
"These are first responders, these are people trying to protect us, they also need to be protected," Nakawatase said. "Creating that bond ... with the firefighters, that brings another level of connection within the community, and it also strengthens the community's emergency response team as a whole."
The team also works on projects involving metabolic, mental and cardiovascular health, among other topics. Ghasemzadeh said EMIL will continue its work to create health systems with broad applications, utilization options and public confidence.
"How do you scale technology?" Ghasemzadeh said. "You want to make sure that you make it robust. You want to make sure you make it transparent. You want to make sure you increase trust in these systems in general."
Edited by Kate Gore, Senna James and Pippa Fung.
Reach the reporters at dforres5@asu.edu.
Like The State Press on Facebook and follow @statepress on X.
