UNIVERSITY PARK, PA – Wearable medical sensors have opened the door to remote health monitoring and treatment evaluation. But making a diagnosis and treatment plan based on multiple data points – such as muscle movement, heart rate, breathing or speaking and swallowing – can be difficult for health care providers to act quickly.
As reported in Nature Communications, engineering researchers have created a machine learning platform that can more efficiently analyze and predict data points collected by wearables. They applied the platform to a new flexible, wearable neck sensor that records vibrations and electrical muscle impulses from the neck area to monitor a user’s speech and swallowing patterns.
James L. Henderson’s principal investigator Huanyu “Larry” Cheng said, “The health care market is focused on developing a soft, stretchy device over the throat to continuously monitor the muscles and swallowing movements of patients with throat conditions and to appropriately diagnose and treat them.” equipment is required.” , Junior Memorial Associate Professor of Engineering Science and Mechanics at Penn State. “The devices currently in use provide only limited information about a patient’s health status and are cumbersome and inconvenient.”
According to Cheng, the wearable patch is made of a composite hydrogel electrode interface, which was designed to withstand the user’s movements on the skin surface while maintaining good signal quality. Insoluble hydrogel materials are flexible and easy to apply and remove.
“The hydrogel can directly contact the skin and provide a conductive property for the sensor,” Cheng said.
According to co-principal investigator Libo Gao, an associate professor at the Pen-Tung Sah Institute of Micro-Nano Science and Technology at Xiamen University in China, the hydrogel sensor collects vibration and muscle movement data to feed machine learning algorithms for analysis. . After collection and data analysis, the data is sent to a custom-built cloud interface, where health care providers can access it remotely.
“Patient data is collected by the patch at different frequencies, depending on the statistical type, such as swallowing, speaking or breathing,” Gao said. “The algorithm groups the four frequencies into a streamlined output, making it more useful for health care providers to quickly view the data and make decisions.”
According to Gao, the algorithm has adaptive capabilities and memory functions, meaning that after collecting a patient’s neck movement data for one minute and training offline for three hours, it can recognize patient data with more than 90% accuracy. Can predict.
Clinicians can use predictive data to inform earlier diagnoses, Cheng said, as well as predict how treatments might work.
“Adaptive machine learning can also help take into account individual differences in data across larger populations, so researchers can make inferences about the health of larger populations based on individual datasets,” Cheng said.
In addition to Cheng and Gao, co-authors include Hongcheng Xu, Weihao Zheng, Weidong Wang, Yangbo Yuan, Ji Zhang, Zimin Huo, Ningjuan Zhao, Yuxin Qin, Ke Liu, Ruida Xi, Gang Chen, Haiyan Zhang, Chu Tang and Junyu. Are included. Yan from Xidian University, China; Yang Zhang, Daqing Zhao and Lu Wang from Air Force Medical University, China; Yunlong Zhao from Xiamen University, China; Yujiao Wang from Tsinghua University, China; Qi Ge from Southern University of Science and Technology, China; and Yang Lu from the University of Hong Kong, Hong Kong SAR.
National Natural Science Foundation of China, Natural Science Foundation of Shaanxi Province, China Postdoctoral Science Foundation, Fundamental Research Fund for Central Universities, Key Research and Development Program of Shaanxi, Shenzhen-Hong Kong-Macao Technology Research Program, Fundamental Research Fund for Central Universities Fund and the Innovation Fund of Zidian University funded this work.
