When the COVID-19 outbreak first started in early 2020, no one, even the scientists, knew much about the virus and how it was transmitted. To analyse the transmission patterns of the disease among different populations, Dr. Yang Liu, Assistant Professor of the Department of Computer Science, HKBU, and his research team quickly initiated a modelling study to examine the spread of the disease through computational modelling and AI technology.
Launched in February 2020, the study is one of the first modelling studies on COVID-19 in the world. Dr. Liu and his team carried out the study by discussing with the experts from China CDC (Chinese Center for Disease Control and Prevention) and using the data collecting from governmental websites, such as The Health Commission of Hubei Province and the Beijing Municipal Health Commission, and developed a computational method to measure the contact intensity between people from spatial and temporal perspectives.
“Different from many existing disease modellings which focus on predicting the total number of cases in the future, what we are exploring is the intrinsic transmission pattern of the disease,” says Dr. Liu. The modelling can also simulate the situations when different intervention strategies applied. “With such simulated data, the public health authorities and decision makers can choose the most appropriate strategy to control the disease or stop its spread.”
Figure 1: Measurement of the intensity of social contacts among seven age-groups (G1: 0-6; G2: 7-14; G3: 15-17; G4: 18-22; G5: 23-44; G6: 45-64; and G7: 65 or above) in four major settings: (A) households; (B) schools; (C) workplaces; and (D) public/community. Addressing this key question enables us to gain insights into the retrospective and prospective situations of the disease outbreak; this in turn will help further answer a series of questions in control and prevention of the disease; namely, how future risks and trends in different regions may evolve, how effective different intervention strategies can be in controlling the outbreak, and what may happen if people gradually return to schools and workplaces in the later stage of the outbreak at some point.
A Timely Tool for COVID-19 Intervention Planning
The result of their study was published on The Lancet’s EClinialMedicine, being the first paper ever published on the journal by a research team from the city and one of the most cited articles on the journal in the past year. The study was later generalized and deepened using the data from Hong Kong. “We are using the experience that we have learned in Hong Kong, to adapt and enhance the model for some other cities and countries, such as the US, Canada, Japan and some European countries,” says Dr. Liu.
Figure 2: The cross-district and cross-population transmission patterns throughout different waves of COVID-19 outbreaks in Hong Kong between January 23, 2020, and January 8, 2021. By revealing transmission patterns using the proposed spatiotemporal connectivity analysis, we gained insights into the dynamism and heterogeneity of COVID-19 transmission, and into the optimization and implementation of common pharmaceutical (e.g., the vaccination of high-risk populations) and nonpharmaceutical (e.g., the lockdown of high-risk areas) intervention strategies.
The codes and programme of the modelling are also available for people to access and download. “If there’s any researcher wants to repeat our analysis or input their own data for similar analysis, they can do that,” he says.
Computational epidemiology and infectious disease modelling for prevention and control have been Dr. Liu’s research interests. “My mother is a doctor in a mainland hospital. That’s why I’ve been always interested in health-related issues, people and stories. And
my father is an engineer. Born in such a science-oriented family, I want to know why about everything. And the best way to know why is conducting scientific research,” says Dr. Liu, who was a Post-Doctoral Research Associate with the Department of Statistics at Yale University and a visiting scholar of the Robotics Institute at Carnegie Mellon University.
Knowledge Transfer for Malaria Control and Prevention
Since 2018, Dr. Liu has been engaged in research in malaria control and prevention in China and Southeast Asia. The research project, first started back in 2010, aims to utilise AI and machine learning technology to help low-income countries and remote areas to solve their malaria infection problems.
Tengchong county, which used to be one of the most malaria-infected areas in the China-Myanmar border, was the area that Dr. Liu worked on. “Originally, we thought that it was just another regular data mining or data feeding problem. After we visited the area, we found that one of the most critical challenges was that there were more than 200 villages in the county. From village to village, it generally takes about two hours by car. But there were only four to five CDC staff who possessed the know-how to perform complete disease diagnosis, case investigation, and epidemic treatment,” he recalls.
Figure 3: Dr. Liu and his team carried out field studies in the China-Myanmar border for malaria control and prevention. (a) In December 2016, Dr. Liu (fourth from right in the front row), together with the public-health experts from NIPD, Tengchong CDC, and HPA, visited a villager in Diantan town, Tengchong city, who have had malaria when she was in Myanmar. (b) In March 2019, Dr. Liu (first from right) visited a villager in Dong village, Tengchong city, who have had malaria when he was in Myanmar. (c) In April 2019, Dr. Liu’s PhD student, Liu Mutong, trained the Tengchong CDC staff to use the case detection system developed by Dr. Liu’s team. (d) In April 2019, Dr. Liu (second from left) taught the village doctor in the health center of Xinzhai village, Mang city to use our developed mobile application to screen high-risk population.
The AI-enabled infectious disease control and prevention developed by Dr. Liu’s team estimated the spatio-temporal mobility patterns of the people, enabling the public health authority to decide which were the key places that they should allocate their anti-malaria resources. Tengchong later successfully achieved malaria elimination and became one of the most successful examples.
Currently, Dr. Liu is working with Ministry of Health of Cambodia to apply new AI technologies to assess the malaria transmission risks in six provinces in Cambodia. “Compared to Tengchong, there are more forest in Cambodia. The patterns of human mobility as well as the receptivity of the area are totally different,” he says. “When we are building the model, we have to take that kind of spatial and temporal heterogeneity into consideration to make the model work.”
New Computer Science Technologies to Solve Real World Problems
With more public health institutions intensively collaborating with different kinds of modelling groups, computer science is surely playing a more significant role in today’s public health system. “For human being, we have limited capacity to memorise all the data, but if we use the computer and design the program properly, then we will have the capacity to analyse the data for us,” says Dr. Liu. By using the historical data together with the knowledge of the domain experts, he explains, computer scientists help to make scientific grounded and accurate analysis in advance, so that interventions can be carried out in a timelier manner.
“I really hope that my research results and outputs can be implemented and deployed by other domain experts or authorities as their guidelines to solve real world problems,” says Dr. Liu. “That’s what we are doing, collaborating with the China CDC, WHO, NGOs, ministry of health in Southeast Asia and African countries. I’m trying to make my research and scientific discoveries socially impactful and useful.”
Figure 4: A novel machine learning model inspired by the real-world challenges of data complexity --- An Interactively- and Integratively-connected Deep Recurrent Neural Network (I2DRNN) model with Information-Theoretic Guarantee was proposed, aiming at characterizing the intrinsic, multi-scale dependency among spatiotemporal data in various real-world tasks such as disease prediction, climate forecast, and traffic prediction.
Related Articles
Yang Liu, Zhonglei Gu, Shang Xia, Benyun Shi, Xiao-Nong Zhou, Yong Shi, and Jiming Liu, What are the underlying transmission patterns of COVID-19 outbreak? – An age-specific social contact characterization, EClinicalMedicine, The Lancet Journals, 22, 100354, May, 2020.
Yang Liu, Zhonglei Gu, and Jiming Liu, Uncovering transmission patterns of COVID-19 outbreaks: A region-wide comprehensive retrospective study in Hong Kong, EClinicalMedicine, The Lancet Journals, 36, 100929, June, 2021.
Qi Tan, Yang Liu, and Jiming Liu, Demystifying Deep Learning in Predictive Spatiotemporal Analytics: An Information-Theoretic Framework, IEEE Transactions on Neural Networks and Learning Systems, 32(8): 3538-3552, 2021.
Figure 1: Measurement of the intensity of social contacts among seven age-groups (G1: 0-6; G2: 7-14; G3: 15-17; G4: 18-22; G5: 23-44; G6: 45-64; and G7: 65 or above) in four major settings: (A) households; (B) schools; (C) workplaces; and (D) public/community. Addressing this key question enables us to gain insights into the retrospective and prospective situations of the disease outbreak; this in turn will help further answer a series of questions in control and prevention of the disease; namely, how future risks and trends in different regions may evolve, how effective different intervention strategies can be in controlling the outbreak, and what may happen if people gradually return to schools and workplaces in the later stage of the outbreak at some point.
Figure 2: The cross-district and cross-population transmission patterns throughout different waves of COVID-19 outbreaks in Hong Kong between January 23, 2020, and January 8, 2021. By revealing transmission patterns using the proposed spatiotemporal connectivity analysis, we gained insights into the dynamism and heterogeneity of COVID-19 transmission, and into the optimization and implementation of common pharmaceutical (e.g., the vaccination of high-risk populations) and nonpharmaceutical (e.g., the lockdown of high-risk areas) intervention strategies.
Figure 3: Dr. Liu and his team carried out field studies in the China-Myanmar border for malaria control and prevention. (a) In December 2016, Dr. Liu (fourth from right in the front row), together with the public-health experts from NIPD, Tengchong CDC, and HPA, visited a villager in Diantan town, Tengchong city, who have had malaria when she was in Myanmar. (b) In March 2019, Dr. Liu (first from right) visited a villager in Dong village, Tengchong city, who have had malaria when he was in Myanmar. (c) In April 2019, Dr. Liu’s PhD student, Liu Mutong, trained the Tengchong CDC staff to use the case detection system developed by Dr. Liu’s team. (d) In April 2019, Dr. Liu (second from left) taught the village doctor in the health center of Xinzhai village, Mang city to use our developed mobile application to screen high-risk population.
Figure 4: A novel machine learning model inspired by the real-world challenges of data complexity --- An Interactively- and Integratively-connected Deep Recurrent Neural Network (I2DRNN) model with Information-Theoretic Guarantee was proposed, aiming at characterizing the intrinsic, multi-scale dependency among spatiotemporal data in various real-world tasks such as disease prediction, climate forecast, and traffic prediction.
