"In the study, we screened out several at-risk babies through the calibration of professional doctors, and then contacted the parents of the babies through the doctors and recommended that they go to a higher hospital for a comprehensive examination. In this moment, we truly feel the importance of what we are doing. ”
Referring to a recent study he participated in, Dr. Zhang Senhao and Dr. Bao Benkun from the Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, still find it very meaningful.
Figure | Zhang Senhao (Source: Zhang Senhao)
Recently, he and his collaborators jointly created a flexible sparse sensor network system that can achieve ultra-high accuracy and automatic classification in the evaluation of restless movements of infants.
At present, the results of experiments conducted by the team and clinical hospitals show that this technology can quickly and effectively carry out large-scale rapid screening of neonatal cerebral palsy risk.
Within a few years, this technology will be able to be rolled out to more regions and become a "vaccine" similar to a "vaccine" for newborns.
（来源：Advanced Science）
Due to the characteristics of low cost and low resource dependence, this technology can still be effectively operated even in areas with underdeveloped medical conditions such as the central and western regions, and is expected to greatly promote the development of maternal and child health in the mainland.
"In particular, if it can be popularized in areas with low medical standards, and the neurodevelopmental and behavioral habits of newborns can be screened, intervened and rehabilitated as soon as possible, the burden on families can be better reduced. Zhang Senhao said.
（来源：Advanced Science）
日前，相关论文以《用于新生儿不安运动自动早期评估的柔性智能稀疏传感器网络》（Intelligence Sparse Sensor Network for Automatic Early Evaluation of General Movements in Infants）为题发在 Advanced Science（IF 15.1）。
Dr. Benkun Bao and Dr. Senhao Zhang from the Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, are working together.
Senhao Zhang, Prof. Hongbo Yang and Prof. Xiankai Cheng from the Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of Sciences, and Prof. Huanyu Cheng from Pennsylvania State University served as co-corresponding authors [1].
Figure | Related papers (Source: Advanced Science)
According to reports, neonatal cerebral palsy screening has always been an area of concern for clinical research.
However, the previous clinical evaluation method mainly relied on experienced professionals to assess whether the newborn has a risk of cerebral palsy by observing the spontaneous movement patterns over a period of time, and then confirm whether further imaging tests are needed.
It is undeniable that this kind of neurodevelopmental assessment through neonatal voluntary movement has a certain convenience and potential value for large-scale screening.
However, the lack of professional doctors limits its further promotion, especially in areas with low medical resources.
Zhang Senhao said: "This project comes from the medical-engineering integration project jointly participated by researcher Cheng Xiankai and researcher Yang Hongbo of our institute and the First Affiliated Hospital of Jilin University. ”
This project aims to promote a number of potential front-line medical-engineering integration and interdisciplinary projects to solve clinical pain points and difficult problems, and this project is fortunate to receive support.
At the beginning, the research group did not have a good solution in terms of wearable mode and data stability.
With the support of Prof. Hongbo Yang's International Partnership Program, Senhao Zhang began to participate in the joint doctoral training in Prof. Huanyu Cheng's group at The Pennsylvania State University (PSU) in 2020.
During this discussion, they concluded that the use of flexible structures and the use of material design methods should be more suitable for this clinical application scenario.
"Therefore, we started this research with the PSU team, and under the guidance of Professor Cheng Huanyu, we began to design the structure and materials, and optimized them for the networking method and network stability," said Zhang Senhao. ”
It is understood that most of the previous studies have captured the movement through the "video method" and then digitized.
However, this method does not guarantee privacy, and the action recognition is susceptible to environmental interference, and the large amount of background noise will make the data volume extremely large, which limits the development in clinical practice.
Based on this, the research group hypothesizes: Is it possible to complete the digitization of neonatal movements through direct movement information collection?
Considering the fragility of the newborn's skin, they completed the flexible design of the sensing node through the "island-bridge" structure. At the same time, the use of highly biocompatible materials further guarantees the skin safety of newborns.
In order not to interfere with the voluntary movement of the newborn by the wearing of the sensor, the number of sensing nodes needs to be reduced as much as possible.
Through algorithm optimization and other means, they only arranged sensors in five places on the limbs and head of the newborn, and realized the construction of a sparse sensor network through Bluetooth Low Energy.
（来源：Advanced Science）
Once the system design was stable, they began to initiate clinical demonstration studies. With the support of the First Affiliated Hospital of Jilin University, Suzhou Children's Hospital, and Quwo County Hospital of Traditional Chinese Medicine in Shanxi Province, they successfully obtained a batch of clinical data.
After preliminary analysis of the data, they found significant differences between children at risk and those at risk in the time and frequency domains. However, there is still no direct threshold that can be used as a criterion.
Specifically, they found that relying on traditional classification methods could not achieve high-accuracy screening, so they began to use machine learning to build classification models.
However, the overall model should not be too large, otherwise the demand for computing power will increase, so that it cannot be delivered to low-level medical areas.
Therefore, it is their goal to create a classification model that is as lightweight and miniaturized as possible. At the same time, they don't want the algorithm they are building to be a "black box" algorithm.
Therefore, the researchers tried to start from the clinician's diagnostic criteria as much as possible to summarize the extraction of eigenvalues.
At the same time, the resulting eigenvalues are ranked according to the degree of relevance, so as to identify those eigenvalues that are more effective for classification and ensure that these eigenvalues can be clinically interpreted.
Only in this way can the dimension of eigenvalues be further reduced, so as to create a lightweight algorithm model.
In the process of building the model, considering that it must occupy as little computing resources as possible, Dr. Bao Benkun of the team optimized the dimensionality reduction and algorithm redundancy of the feature dimension in order to obtain a small algorithm model with high accuracy.
With the help of artificial intelligence technology, they successfully developed the required classification model. Moreover, as the amount of data increases, this data processing method can also be optimized, so as to achieve a lighter, faster, and more efficient classification algorithm.
In order to promote the application in low-level medical resource areas, it is necessary to develop low-cost hardware systems and miniaturized automatic identification algorithms that are easy to deploy.
In the end, thanks to their efforts, the cost of the entire system was less than 500 yuan. At the same time, by reducing the number of eigenvalues and using logistic regression algorithm, the research group has achieved the construction of the minimum model under the premise of a high recognition rate of 99.9%.
It is also reported that this flexible physiological sensing network can not only be used for neonatal restless motion detection, but also can achieve more physiological information acquisition, so as to be used in other clinical aspects.
For example, accelerometers can be used to collect information such as heart rate and respiration rate, as well as swallowing ability, mobility, etc., which can be used in the ICU intensive care unit.
On the other hand, for rapid screening of neonatal cerebral palsy, restless movement monitoring is mainly indicated for neonates up to 20 weeks.
For older infants who are already able to crawl, they are also analysing crawling behaviour and cerebral palsy so that they can be used to assess their neurodevelopmental abilities in older infants.
In terms of application, they are also actively contacting some local maternal and child health stations in the central and western regions to better promote the system.
At the same time, they are also actively trying to see if it is possible to build a battery-free sensing system.
If radio frequency can be used for wireless power supply, the weight of the sensor can be further reduced by removing the battery module, thereby reducing the impact of wearing the sensor on the spontaneous movement of the newborn.
Resources:
1.Bao, B., Zhang, S., Li, H., Cui, W., Guo, K., Zhang, Y., ... & Cheng, H. (2024). Intelligence Sparse Sensor Network for Automatic Early Evaluation of General Movements in Infants. Advanced Science, 2306025.
Operation/Typesetting: He Chenlong