Addressing COVID-19 Challenges

Find FAQ and related info pages on our COVID-19 Response Page

The following is a list of USC Viterbi Researchers working on COVID-19 related projects. Please click the individual’s / lab’s name to jump to their section.



Solutions for PPEs – Armani Lab - Andrea Armani & Rosemary She (Keck School of Medicine of USC)

Build-at-home UV-C disinfection system for healthcare settings

Significant research has shown that UV-C exposure is an effective disinfectant for a range of bacteria and viruses, including coronaviruses. As such, a UV-C treatment in combination with a chemical wipe, such as EPA hydrogen peroxide, is a common cleaning protocol in a medical setting, and such disinfection protocols have gained in importance during the current COVID-19 pandemic due to the need to reuse PPE. However, given the substantial increase in patient volume, the quantity of materials requiring disinfection exceeds the UV-C equipment throughput capabilities at many medical facilities. Therefore, there is a need for a UV-C disinfection system that can be rapidly deployed. In response to this demand, we designed, constructed, and validated a UV-C disinfection system from readily accessible components; specifically, a plastic bin, UV-C light bulb and conventional light housing. To further improve the performance, the interior of the tub was spray-painted with chrome paint, forming a low quality-factor (Q) fabry-perot optical cavity. As part of this work, a set of modular design criteria which allows for flexibility in component selection without degradation of UV-C dose performance is established. This flexibility is critical given the current fluctuating availability of source materials. The disinfection capabilities of the system are validated using Bacillus cereus, a gram-positive endospore-forming bacteria.

Using Emergency Ventilator Analysis, Design, Manufacturing as Case Studies to Enhance Student Learning Experiences - Ivan Bermejo-Moreno (USC AME), SK Gupta (USC AME), Alec Kanyuck (USC AME), Mitul Luhar (USC AME), Anita Penkova (USC AME) and Daniel Stemen (USC Keck)

Ventilator ImageUnder the COVID-19 pandemic, a shortage of mechanical ventilators has been identified as a limiting factor in the ability of health care systems worldwide to provide treatment to patients whose condition evolves into acute respiratory distress syndrome (ARDS).

There has been significant interest from the community to develop emergency ventilators and many different open-source designs have been proposed.  There is substantial debate about the effectiveness of these designs.  Many students at USC are interested in joining efforts to mitigate COVID-19 related challenges.  We believe that emergency ventilators present a very useful case study for enhancing student learning experiences at this time of crisis.

We are identifying possible lines of work that could be relevant in our community to address the critical shortage of ventilators and related equipment by establishing collaborations between academic institutions, medical centers, local industry partners, and government authorities. The objective is to complement ongoing, large-scale efforts by government and industry, synergizing expertise in respiratory medical care with academic and industrial expertise in mechanical engineering (design, advanced manufacturing, robotics, controls), and leveraging unique fabrication facilities for rapid prototyping (including 3D printers, laser cutters, mills, machine shop equipment, etc.).

We have identified eight short-term projects that can be accomplished by small groups to improve existing open-source ventilator designs (mechanical and electronic components), their manufacturing processes (materials, automation, assembly), and production/delivery considering supply chains, component outsourcing, transportation network, and rapid delivery. AME 546 students are currently working on improving emergency ventilator designs from the manufacturing cost perspective. We plan to engage additional students during the summer to expand student research activities in this area.


Deactivation of Airborne and Surface-Borne Viruses using Nanosecond Pulsed Plasmas – Cronin Lab – Steve Cronin

We are developing a based plasma reactor system capable of treating air handling systems in ships, aircraft, hospitals, and submarines to deactivate viruses and other biological agents.1, 2Our technology is based on a transient pulsed plasma that is generated by a USC-patented nanosecond high voltage pulse generator, which consumes far less energy than conventional RF plasma sources. The transient nature of the plasma necessitates that very little current is drawn in creating the plasma. That is, once the streamer is created, the applied field collapses before a substantial amount of current (and hence electric power) can flow. Because of its transient nature, this is a cold plasma, in which the electron energies ~10’s of eV (T=104K), while the vibrational modes of the molecules remain at room temperature. These “hot” electrons enable effective chemical pathways in the remediation of biological agents through the generation of reactive oxygen species (ROS).

We are also developing plasma jets capable of cleaning surfaces 100mm×100mm areas in less than 5 seconds for the controlled mitigation of chemical and biological agents, including COVID-19. These plasma jets have a uniquely long “throw distance” not achievable with conventional RF power supplies, and maintain low temperatures for treating sensitive surfaces. In our proposed work, we will partner with Prof. Pin Wang and Transient Plasma Systems, Inc. to demonstrate and optimize the efficacy of this approach for deactivating airborne and surface-borne viruses using this plasma-based approach.

Cronin Covid research

Collaborators:

  • Transient Plasma Systems, Inc.
  • Professor Pin Wang

Link:

https://cronin-lab.usc.edu/publications/


Online impact of COVID – Machine INtelligence and Data Science (MINDS) Research Group – Emilio Ferrara

COVID-19 Social Media Watch

With people moving out of public physical spaces due to isolation and social distancing measures, online platforms became even more prominent tools to track society and studying social media can be more informative than ever to understand how we are collectively coping with this unprecedented crisis. We have been continuously tracking online chatter regarding COVID-19 since mid-January 2020. Our primary focus has been Twitter. The collected dataset comprises of dozens of million tweets about the outbreak and related events worldwide, spanning over two months so far. Data access is the first barrier to research. Therefore, we decided to make our dataset public to foster research on this important problem. With this contribution, our goal is to enable researchers in computational and social sciences and provide them with a shared framework and dataset to study the important issues revolving around COVID-19 and social media discussions. Issues such as misinformation, fear, crisis dynamics, and much more, could be studied thanks to our data collection.
Dataset: https://github.com/echen102/COVID-19-TweetIDs
Paper: https://arxiv.org/abs/2003.07372

Mis- and Disinformation in COVID-19 online discussion

The problem of misinformation and disinformation surrounding the ongoing outbreak is as critical, in our opinion, as the pandemic itself. It can have consequences that go well beyond public health and reach into the economy, public safety, political and policy issues. We have been compiling an ongoing list of references tracking reports of mis- and disinformation, including foreign sponsored operations carried out by China, Russia, and others. We are in the process of putting together technological safeguards to automatically detect and characterize such efforts. The initial focus will be on the identification of social media bots, automated accounts typically used to push narratives and carry out coordinated influence operations. We will then move to the more complex task of uncovering trolls, social media human users pushing an agenda, typically as part of state-sponsored interference operations.

Local and Global Emotion Dynamics (fear, anger, hate speech)

How do local fear and emotional swings precipitate and contribute to the spread of global fear and panic? Can we classify different kinds of public responses to a pandemic through social media cues? In particular, do emotion trends (e.g. hate speech, fear) change with outbreaks and major events?  What kinds of emotional contagions are predictive of the development of a local fear epidemic into a global pandemic? How do these emotions facilitate their own spread within a community and across borders and cultures? What features of the social media environment trigger certain emotional responses?

We are using the collected Twitter data as a proxy for public reaction and discourse related to COVID19. We have been working on defining “local” tweets by geolocating tweets (both through the explicit geolocation data provided by Twitter and user-supplied location data), have been working on classifying per Tweet emotions, and applying hate speech detection methods to the Tweets. Our future plans include translating tweets from different languages in order to detect emotions and responses to emotions in different cultures (language as a proxy).

Predictability Limits of Real-Time Outbreak Surveillance

The number of Coronavirus cases is growing all around the world, yet, data is vastly underreported due to limited tests, asymptomatic or mild cases. Accounting for underreporting of cases is crucial to improve the reaction times to disease outbreaks, yielding accurate and informative information about temporal dynamics as early as possible. Our past work proposes a framework to study dynamic phenomena in partially observed systems such as infectious disease outbreaks. For the current Coronavirus outbreak, we are applying our framework to create better assessment tools and develop methodologies to get more accurate estimates and predictions of the dynamics of the outbreak.

Collaborators: Kristina Lerman (USC/ISI)


Semi-Autonomous Robots for Disinfection and Sterilization - USC Viterbi Center for Advanced Manufacturing (CAM) - S.K. Gupta and Andrea Armani

Traditional methods for performing disinfection and sterilization tasks require significant human effort when dealing with cluttered space or manufacturing shop floors. Moreover, the availability of properly trained human workers is in short supply during pandemics. Current methods for disinfection and sterilization efforts are expected to be very slow and cannot be rapidly scaled.

Robots can be very useful in disinfecting and sterilizing surfaces. We are interested in developing a robotic platform to meet the following goals: (1) improving human productivity, (2) reducing risks to human workers, (3) eliminating human errors during disinfection and sterilization. We are developing a mobile manipulator operating under a remotely located human supervisor to conduct disinfection operation. Our method has potential to dramatically reduce the need for humans to be in close vicinity of the contaminated area. Human operators can task the robot and monitor its progress from a safe distance. The use of a mobile manipulator will also dramatically reduce the amount of physical work performed by humans.

GuptaRobot

Our preliminary prototype is called ADAMMS-UV (Agile Dexterous Autonomous Mobile Manipulation System-UV), a robot to perform disinfection tasks in public spaces such as offices, labs, schools, hotels, and dorms using UV light. The UV light mounted robots currently in markets have a UV column on a mobile base that goes around in rooms and stands still at certain places in the room so that surfaces that are exposed to the light are disinfected. ADAMMS-UV is a semi-autonomous mobile manipulator that uses a UV light wand mounted on a robotic arm to reach spaces that cannot be treated by such UV source mounted robots. In addition, it also has a large UV light source to disinfect large open spaces. It can use the gripper to open drawers, closets and manipulate objects to perform a detailed sanitization on hard to reach surfaces. UV light is a proven disinfectant. Coronavirus on a surface can be killed when exposed to UV light of appropriate intensity for a sufficient amount of time. ADAMMS-UV can hold a UV wand over a surface and move it the right speed. The robot can do this task consistently without making any mistake. This task will be physically taxing and risky for humans.

ADAMMS-UV is controlled by a remote operator located far away from the risk zone. The remote operator can provide a goal position for the mobile base to go to or drive it manually. Based on the user instructions, the robotic arm can then scan the surroundings using a stereo camera and build a 3D scene. The operator then provides the robot with high-level instructions of what area to disinfect or what object to manipulate by selecting in the 3D scene via the GUI developed by us. The robot then computes the motion plans for the arm at the backend and shows a preview to the user, who can then either give a go-ahead to the robot or discard the motion and ask it to plan again. The user can also manually modify the motions of the arm to accurately manipulate objects. This greatly reduces the human effort to operate the robot and one operator can operate multiple such robots simultaneously. With such a robot operational in public spaces, we believe that we can disinfect large surfaces with the UV light column on the robot and use the robotic arm to disinfect occluded surfaces.

Link: https://youtu.be/-YdsP9M-Luw

Team:

Hyojeong Kim, Pradeep Rajendran, Shantanu Thakar, Sarah Al-Hussaini, Rex Jomy Joseph, Alec Kanyuck, Rishi Malhan, Omey Manyar, Brual C. Shah, Ariyan M. Kabir, Andrea Armani, and Satyandra K. Gupta

Collaborators:

Jeremy Marvel


A Predictive Model for Planning COVID-19 Treatment - Computation and Data Driven Discovery Group – Assad Oberai

Of all the COVID-19 patients that are hospitalized, around 1/3 require intensive care. The need for intensive care is brought on by respiratory distress, secondary sepsis and onset of multiorgan failure. It is believed that several clinical markers like lymphocyte count, d-dimer levels, temperature, SOFA score, and x-ray imaging markers like the development of bilaterally multifocal consolidations, are indicative of the potential need for intensive care. However, currently there is no quantitative criterion that can accurately predict whether a given COVID-19 patient will require intensive care at some point. The development of such a criterion or a predictive modeling system will aid in triaging COVID-19 patients as they enter a care facility into a cohort that is likely to require intensive care and another that is not. It will therefore permit a more nuanced use of resources and care at a time when both are scarce. Such a system will also allow for a more accurate prediction of the demand for intensive care resources at a given site.

We are developing a predictive modeling system that can accurately predict the likelihood of a covid-19 patient requiring intensive care. The input to this system will be clinical data that is recorded when the patient is admitted, and semi-quantitative and quantitative information gathered from x-ray images of the chest. This data will be used in unsupervised and supervised machine learning algorithms to compute this likelihood as well the features in the dataset that contribute most strongly towards the prediction. The feasibility of our proposed classification system is supported by evidence that has emerged linking certain clinical and imaging features to the ultimate severity of this disease. Thus, training a machine learning algorithm on this data is likely to uncover the underlying highly non-linear correlations between the clinical and imaging features and the likelihood of intensive care.

Collaborators:
Assad A Oberai (USC Viterbi), Vinay Duddalwar, Ali Gholamrezanezhad, Bhushan Desai, Bino Varghese, Neha Nanda (USC Keck)


@CoronaSurveys: Monitoring the Incidence of COVID-19 via Open Surveys -  Signal Transformation, Analysis and Compression Group – Antonio Ortega

With several collaborators I am investigating methods to estimate the total number of cases of COVID-19, beyond those that have been confirmed via testing. Since tests are not widely available, often mild or asymptomatic cases are not recorded. Having additional estimates of possible cases can lead to improved estimates of case fatality rates (CFRs). Also, since these surveys are run on a continuous basis, they can help model the temporal evolution of the outbreak and provide some evidence about whether quarantine, social distancing and other measures are being successful.

We have been collaborating with an international team (Europe, AsiaPacific and the Americas) to develop a complete system for continuous polling, to improve our communication (through a new website) and to improve our analysis tools.  The data being acquired is being made publicly available to other researchers. A unique characteristic of these surveys is that they request information about people known to each of the respondents. That way, with even a relatively small number of responses we get indirect information about a much larger number of people.

The current site is: http://coronasurveys.org/
Code and data can be found here: https://github.com/gcgimdea/coronasurveys

Covid Estimate

Our contribution to the CoronaSurveys project has been focused in two areas: i) supporting the new infrastructure, and in particular continuous data collection and improved communication of our results, and ii) mathematical techniques to improve the estimates. In the latter topic we are building on our expertise on graph signal processing. We view the data being collected as samples on a social graph: each user responding to the survey provides information about "neighboring nodes" on this social graph. We are currently doing preliminary work on methods that can improve estimates by taking into account the implied graph connectivity. We are also working with the team to modify the questionnaires to improve estimates while maintaining complete privacy.

Collaborators:
USC: Dr Benjamin Girault
International collaborators: Updated list can be found here: https://coronasurveys.org/info/team/


Online impact of COVID – Machine INtelligence and Data Science (MINDS) Research Group – Emilio Ferrara

COVID-19 Social Media Watch

With people moving out of public physical spaces due to isolation and social distancing measures, online platforms became even more prominent tools to track society and studying social media can be more informative than ever to understand how we are collectively coping with this unprecedented crisis. We have been continuously tracking online chatter regarding COVID-19 since mid-January 2020. Our primary focus has been Twitter. The collected dataset comprises of dozens of million tweets about the outbreak and related events worldwide, spanning over two months so far. Data access is the first barrier to research. Therefore, we decided to make our dataset public to foster research on this important problem. With this contribution, our goal is to enable researchers in computational and social sciences and provide them with a shared framework and dataset to study the important issues revolving around COVID-19 and social media discussions. Issues such as misinformation, fear, crisis dynamics, and much more, could be studied thanks to our data collection.
Dataset: https://github.com/echen102/COVID-19-TweetIDs
Paper: https://arxiv.org/abs/2003.07372

Mis- and Disinformation in COVID-19 online discussion

The problem of misinformation and disinformation surrounding the ongoing outbreak is as critical, in our opinion, as the pandemic itself. It can have consequences that go well beyond public health and reach into the economy, public safety, political and policy issues. We have been compiling an ongoing list of references tracking reports of mis- and disinformation, including foreign sponsored operations carried out by China, Russia, and others. We are in the process of putting together technological safeguards to automatically detect and characterize such efforts. The initial focus will be on the identification of social media bots, automated accounts typically used to push narratives and carry out coordinated influence operations. We will then move to the more complex task of uncovering trolls, social media human users pushing an agenda, typically as part of state-sponsored interference operations.

Local and Global Emotion Dynamics (fear, anger, hate speech)

How do local fear and emotional swings precipitate and contribute to the spread of global fear and panic? Can we classify different kinds of public responses to a pandemic through social media cues? In particular, do emotion trends (e.g. hate speech, fear) change with outbreaks and major events?  What kinds of emotional contagions are predictive of the development of a local fear epidemic into a global pandemic? How do these emotions facilitate their own spread within a community and across borders and cultures? What features of the social media environment trigger certain emotional responses?

We are using the collected Twitter data as a proxy for public reaction and discourse related to COVID19. We have been working on defining “local” tweets by geolocating tweets (both through the explicit geolocation data provided by Twitter and user-supplied location data), have been working on classifying per Tweet emotions, and applying hate speech detection methods to the Tweets. Our future plans include translating tweets from different languages in order to detect emotions and responses to emotions in different cultures (language as a proxy).

Predictability Limits of Real-Time Outbreak Surveillance

The number of Coronavirus cases is growing all around the world, yet, data is vastly underreported due to limited tests, asymptomatic or mild cases. Accounting for underreporting of cases is crucial to improve the reaction times to disease outbreaks, yielding accurate and informative information about temporal dynamics as early as possible. Our past work proposes a framework to study dynamic phenomena in partially observed systems such as infectious disease outbreaks. For the current Coronavirus outbreak, we are applying our framework to create better assessment tools and develop methodologies to get more accurate estimates and predictions of the dynamics of the outbreak.

Collaborators: Kristina Lerman (USC/ISI)


Semi-Autonomous Robots for Disinfection and Sterilization - USC Viterbi Center for Advanced Manufacturing (CAM) - S.K. Gupta and Andrea Armani

Traditional methods for performing disinfection and sterilization tasks require significant human effort when dealing with cluttered space or manufacturing shop floors. Moreover, the availability of properly trained human workers is in short supply during pandemics. Current methods for disinfection and sterilization efforts are expected to be very slow and cannot be rapidly scaled.

Robots can be very useful in disinfecting and sterilizing surfaces. We are interested in developing a robotic platform to meet the following goals: (1) improving human productivity, (2) reducing risks to human workers, (3) eliminating human errors during disinfection and sterilization. We are developing a mobile manipulator operating under a remotely located human supervisor to conduct disinfection operation. Our method has potential to dramatically reduce the need for humans to be in close vicinity of the contaminated area. Human operators can task the robot and monitor its progress from a safe distance. The use of a mobile manipulator will also dramatically reduce the amount of physical work performed by humans.

GuptaRobot

Our preliminary prototype is called ADAMMS-UV (Agile Dexterous Autonomous Mobile Manipulation System-UV), a robot to perform disinfection tasks in public spaces such as offices, labs, schools, hotels, and dorms using UV light. The UV light mounted robots currently in markets have a UV column on a mobile base that goes around in rooms and stands still at certain places in the room so that surfaces that are exposed to the light are disinfected. ADAMMS-UV is a semi-autonomous mobile manipulator that uses a UV light wand mounted on a robotic arm to reach spaces that cannot be treated by such UV source mounted robots. In addition, it also has a large UV light source to disinfect large open spaces. It can use the gripper to open drawers, closets and manipulate objects to perform a detailed sanitization on hard to reach surfaces. UV light is a proven disinfectant. Coronavirus on a surface can be killed when exposed to UV light of appropriate intensity for a sufficient amount of time. ADAMMS-UV can hold a UV wand over a surface and move it the right speed. The robot can do this task consistently without making any mistake. This task will be physically taxing and risky for humans.

ADAMMS-UV is controlled by a remote operator located far away from the risk zone. The remote operator can provide a goal position for the mobile base to go to or drive it manually. Based on the user instructions, the robotic arm can then scan the surroundings using a stereo camera and build a 3D scene. The operator then provides the robot with high-level instructions of what area to disinfect or what object to manipulate by selecting in the 3D scene via the GUI developed by us. The robot then computes the motion plans for the arm at the backend and shows a preview to the user, who can then either give a go-ahead to the robot or discard the motion and ask it to plan again. The user can also manually modify the motions of the arm to accurately manipulate objects. This greatly reduces the human effort to operate the robot and one operator can operate multiple such robots simultaneously. With such a robot operational in public spaces, we believe that we can disinfect large surfaces with the UV light column on the robot and use the robotic arm to disinfect occluded surfaces.

Link: https://youtu.be/-YdsP9M-Luw

Team:

Hyojeong Kim, Pradeep Rajendran, Shantanu Thakar, Sarah Al-Hussaini, Rex Jomy Joseph, Alec Kanyuck, Rishi Malhan, Omey Manyar, Brual C. Shah, Ariyan M. Kabir, Andrea Armani, and Satyandra K. Gupta

Collaborators:

Jeremy Marvel


@CoronaSurveys: Monitoring the Incidence of COVID-19 via Open Surveys -  Signal Transformation, Analysis and Compression Group – Antonio Ortega

With several collaborators I am investigating methods to estimate the total number of cases of COVID-19, beyond those that have been confirmed via testing. Since tests are not widely available, often mild or asymptomatic cases are not recorded. Having additional estimates of possible cases can lead to improved estimates of case fatality rates (CFRs). Also, since these surveys are run on a continuous basis, they can help model the temporal evolution of the outbreak and provide some evidence about whether quarantine, social distancing and other measures are being successful.

We have been collaborating with an international team (Europe, AsiaPacific and the Americas) to develop a complete system for continuous polling, to improve our communication (through a new website) and to improve our analysis tools.  The data being acquired is being made publicly available to other researchers. A unique characteristic of these surveys is that they request information about people known to each of the respondents. That way, with even a relatively small number of responses we get indirect information about a much larger number of people.

The current site is: http://coronasurveys.org/
Code and data can be found here: https://github.com/gcgimdea/coronasurveys

Covid Estimate

Our contribution to the CoronaSurveys project has been focused in two areas: i) supporting the new infrastructure, and in particular continuous data collection and improved communication of our results, and ii) mathematical techniques to improve the estimates. In the latter topic we are building on our expertise on graph signal processing. We view the data being collected as samples on a social graph: each user responding to the survey provides information about "neighboring nodes" on this social graph. We are currently doing preliminary work on methods that can improve estimates by taking into account the implied graph connectivity. We are also working with the team to modify the questionnaires to improve estimates while maintaining complete privacy.

Collaborators:
USC: Dr Benjamin Girault
International collaborators: Updated list can be found here: https://coronasurveys.org/info/team/

@CoronaSurveys: Monitoring the Incidence of COVID-19 via Open Surveys -  Signal Transformation, Analysis and Compression Group – Antonio Ortega

With several collaborators I am investigating methods to estimate the total number of cases of COVID-19, beyond those that have been confirmed via testing. Since tests are not widely available, often mild or asymptomatic cases are not recorded. Having additional estimates of possible cases can lead to improved estimates of case fatality rates (CFRs). Also, since these surveys are run on a continuous basis, they can help model the temporal evolution of the outbreak and provide some evidence about whether quarantine, social distancing and other measures are being successful.

We have been collaborating with an international team (Europe, AsiaPacific and the Americas) to develop a complete system for continuous polling, to improve our communication (through a new website) and to improve our analysis tools.  The data being acquired is being made publicly available to other researchers. A unique characteristic of these surveys is that they request information about people known to each of the respondents. That way, with even a relatively small number of responses we get indirect information about a much larger number of people.

The current site is: http://coronasurveys.org/
Code and data can be found here: https://github.com/gcgimdea/coronasurveys

Covid Estimate

Our contribution to the CoronaSurveys project has been focused in two areas: i) supporting the new infrastructure, and in particular continuous data collection and improved communication of our results, and ii) mathematical techniques to improve the estimates. In the latter topic we are building on our expertise on graph signal processing. We view the data being collected as samples on a social graph: each user responding to the survey provides information about "neighboring nodes" on this social graph. We are currently doing preliminary work on methods that can improve estimates by taking into account the implied graph connectivity. We are also working with the team to modify the questionnaires to improve estimates while maintaining complete privacy.

Collaborators:
USC: Dr Benjamin Girault
International collaborators: Updated list can be found here: https://coronasurveys.org/info/team/


Model for COVID Transmission - FPGA/Parallel Computing Lab - Victor Prasanna

RAPID: ReCOVER -- Accurate Predictions and Resource Allocation for COVID-19 and Epidemics Response

The recent outbreak of COVID-19 and its world-wide impact calls for urgent measures to contain the epidemic. Predicting the speed and severity of infectious diseases like COVID-19 and allocating medical resources appropriately is central to dealing with epidemics. Epidemics like COVID-19 not only affect world-wide health, but also have profound economic and social impact. Containing the epidemic, providing informed predictions and preventing future epidemics is essential for the global population to resume their day-to-day work and travel without fear. Shortage of resources such as masks and testing kits puts undue stress on healthcare system further risking health of the community. Preparedness and better management of available resources would require specific predictions at the level of cities and counties around the world rather than solely at the level of countries. The project ReCOVER will learn infection models for COVID-19 considering the following. (i) Predicting at state/county/city-level rather than country-level as finer granularity is essential in planning and managing resources. (ii) How infectious a person is changes over time. Learning the model through observed data will help in understanding of the temporal nature of the virality. (iii) At such granularity, travel is a significant reason for the spread. (iv) Available data needs to be “corrected” by finding the number of underlying unreported cases that are not observed and yet influence the epidemic dynamics. The project will also solve the resource allocation problem based on the predictions – for instance if a certain number of masks are going to be available next week in a certain state, how should they be distributed across different hospitals in the state (which hospitals and how many in each).

ReCOVER uses a novel fine-grained, heterogeneous infection rate model to perform predictions at various granularities (hospital/airports, city, state, country) while accounting for human mobility. ReCOVER integrates data from various sources to build highly accurate models for prediction of the epidemic across the world at various granularity. The project addresses the issue of unreported cases through modeling it as a hidden state in the probabilistic process. The right granularities of modeling is automatically identified, e.g., when to model a state over its cities to trade-off precision for higher reliability in predictions. The project also formulates and solves a resource allocation problem that can guide the response to contain the epidemic and prevent future outbreaks. This is provided by optimal solutions to resource allocation over a network where each node (representing a region) has a function that captures probabilistic response. While the project obtains data with COVID-19 in consideration, the model and algorithms developed under the project are applicable to a wide class of contagious diseases. The project is also developing an interactive customizable tool that can be used to perform predictions and resource management by a qualified user such as a government entity tasked with managing the epidemic response.

Completed Work: We have already obtained highly accurate models for country-level predictions and state-level predictions in the US. The figure on the right shows the predictions for California. We have demonstrated that human mobility is a key factor in obtaining accurate prediction for some US states. We are currently working on obtaining county-level infection and mobility data to perform more specific predictions.


End-to-End Data System for COVID-19 Spread - Information Laboratory (InfoLAB) - Cyrus Shahabi

An end-to-end system for data acquisition, storage, access, analysis and visualization of COVID-19 spread

It is now common knowledge that identifying infected patients early and then isolating them and all the suspected individuals who may have come into contact with the infected person (aka contact tracing) is of utmost importance for prevention and control of epidemic disease outbreaks such as COVID-19. The traditional method of contact tracing that requires detective work by interviewing patients is too slow and does not scale to a pandemic such as COVID-19. An obvious solution is to use people’s mobile phones to track their locations and then identify nearby people automatically. Some countries such as Israel, South Korea and China have already experimented with this approach to react to COVID-19 [1]. However, there are two main concerns with contact tracing using location data [2]: 1) is it effective enough in identifying suspects, and 2) does it violate personal privacy. In the past several years, my team has developed technologies in the area of spatiotemporal data management and analysis to alleviate both of these concerns. Given this track record, we are now working with three different teams to integrate our technologies in order to utilize location data for better analysis of the spatiotemporal spread of COVID-19. Below I first discuss our technologies addressing the above two concerns and then briefly explain our three collaborative projects in acquiring, analyzing and visualizing COVID-19 mobile data.

Effective Contact Tracing:

A main concern with using mobile data for contact tracing is that due to infrequent or generalized location tracking (e.g., for privacy enhancement or due to device GPS errors), distance may not be accurate enough nor can duration of co-location be captured adequately. Hence, we plan to incorporate the strength of social relationships between two users as an additional factor. Duration of social contact is typically longer, suggesting a higher risk, for dyads with social ties rather than dyads of strangers. This social relationship can be partially explicitly collected from the users or implicitly inferred from their historical trajectories as we have demonstrated in our prior work [3]. The intuition is that if two people have frequent co-locations, especially at not-popular places, it is likely they are socially related.

Privacy-Preserving Contact Tracing:

One way to address the privacy concerns of the virus spread tracking apps is to operate entirely with encrypted data. We have developed a secure location-based alert system that can identify and notify users who were situated in proximity to a specific event of interest - in this case, a patient diagnosed with COVID19. Work on this project has been recently funded by the National Science Foundation. Every mobile user running the app must periodically send their encrypted locations to a central database server. Whenever a patient is diagnosed, a trusted authority, e.g., the CDC, can authorize the creation of a search token based on the movement history of the infected person. The token is a cryptographic object that allows the server to identify and send secure alerts to all users who have been in close proximity to the patient. Through the use of a special kind of encryption, called searchable encryption, the evaluation of the proximity relationship can be performed entirely on encrypted locations. In addition, the system does not learn any information about the locations of any users (including the notified ones), except whether they were in proximity (e.g., within 100 feet) to the diagnosed case or not. This way, a secure alert system for COVID19 could be implemented with minimal intrusion on app users’ privacy. An alternative approach to protecting location privacy is by adding noise to collected locations as we discussed in our prior work [4].

Cyrus research

Project-1:  COVID-19 Mobile Data Collection

In collaboration with colleagues at USC: Peter Kuhn Bhaskar Krishnamachari and Mohamad Naveed, we are developing an app that can be used by students to report their symptoms and corresponding locations. This location dataset can be used in the next project for contact tracing and analysis.

Project-2: COVID-19 Data Management and Analysis

Cyrus 2As part of an invited NSF RAPID Proposal in collaboration with Emory University and UTHealth, we are developing technologies to effectively detect three types of transmission of COVID-19 from mobile data: 1) direct person-to-person transmission, i.e. in close contact with someone infected (simultaneous co-location); 2) fomite transmission, i.e. in contact with a contaminated surface or object at a location visited by someone infected earlier (lagged colocation); and 3) indirect person-to-person transmission by contact with someone who is earlier in contact with someone infected. The first two are rather simple and can be developed as a trajectory range query [5]. The indirect transmission is more challenging both from modeling and computational point of view. We need to find all users that are directly and indirectly in contact with the confirmed case. This can be formulated as spatiotemporal reachability queries, for which we proposed scalable approaches in prior work [6].

Project-3: Map Visualization of COVID-19 Spread

In collaboration with Gabe Kahn at Annenberg School of Communication and Journalism under the famous Crosstown project, we created an online map visualization of COVID-19 spread, see: https://coronavirus.xtown.la/ In sum, the combination of the three projects we are involved with would enable us to create an end-toend system for acquisition, storage, access, analysis and visualization of COVID-19 spatiotemporal data.

References: [1] S. Shekhar and A. H. Shekhar, As COVID-19 Accelerates, Governments Must Harness Mobile Data to Stop Spread, Scientific America Newsletter, March 19, 2020, https://blogs.scientificamerican.com/observations/as-covid19-accelerates-governments-must-harness-mobile-data-to-stop-spread/ [2] Cat Zakrzewski, The Technology 202: We asked more than 100 tech experts if U.S. should use location data to track coronavirus. They were split. PowerPost, March 30, 2020, https://www.washingtonpost.com/news/powerpost/paloma/the-technology-202/2020/03/30/the-technology-202-weasked-more-than-100-tech-experts-if-u-s-should-use-location-data-to-track-coronavirus-they-weresplit/5e80c80b602ff10d49ad757e/ [3] H. Pham, C. Shahabi, and Y. Liu. EBM: an entropy-based model to infer social strength from spatiotemporal data. In Proceedings of the ACM SIGMOD International Conference on Management of Data, SIGMOD 2013, New York, NY, USA, June 22-27, 2013, pages 265–276. ACM, 2013. [4] R. Ahuja, G. Ghinita, and C. Shahabi. A utility-preserving and scalable technique for protecting location data with geo-indistinguishability. In Advances in Database Technology - 22nd International Conference on Extending Database Technology, EDBT 2019, Lisbon, Portugal, March 26-29, 2019, pages 217–228. OpenProceedings.org, 2019. [5] M.-E. Yadamjav, F. Choudhury, Z. Bao, and H. Samet. Efficient Multi-range Query Processing on Trajectories: 37th International Conference, ER 2018, Xi’an, China, October 22–25, 2018, Proceedings, pages 269–285. 09 2018. [6] H. Shirani-Mehr, F. B. Kashani, and C. Shahabi.


Fairness and Efficiency in the Allocation of Scarce Healthcare Resources During the COVID-19 Pandemic - Data-Driven Decision-Making (D3M) Research Group - Phebe Vayanos

One of the major problems that has emerged during the pandemic is that hospitals and states have had to work with more cases than they have resources for (not enough PPE, not enough ventilators, not enough doctors, not enough nurses). This has placed decision-makers in these systems (e.g., doctors and policy-makers) in the extremely difficult position to have to choose who gets access to and priority for these resources. This poses huge ethical dilemmas (need to balance system efficiency with fairness considerations) which are complicated by the diversity of stakeholders in these systems (e.g., doctors, patients, hospitals, governments, rest of the population) and by the high degree of uncertainty under which decisions must be made. Thus, there is an urgent need for disciplined, automated, data-driven approaches for coordinating the allocation of these scarce resources. Through this project, our objective is to design automated prioritization tools for assisting decision-makers in this task to help alleviate some of the pressure they are facing during this pandemic and to ensure that scarce resources are put to best use to maximize system efficiency while ensuring that resources are allocated fairly and in ways that align with stakeholder preferences.

Vayanos Covid-19 Research

Figure 1 legend: Model of the healthcare system during the COVID-19 pandemic as a queuing system. Individuals with different characteristics and conditions arrive over time and are matched to resources that are available (e.g., CCU beds, ventilators). The resource allocation policy determines which patients get matched to what resource and in turn impact the fairness-efficiency characteristics of the policy.

Vayanos Covid-19 Research

Figure 2 legend: Preference elicitation for allocation of COVID-19 resources: we propose to learn the preferences of stakeholders over policy characteristics (e.g., fairness/efficiency trade-off) by asking them pairwise comparisons over policy outcomes.

Link to my homepage: https://sites.google.com/usc.edu/phebevayanos/

Students involved: Caroline M. Johnston (USC Viterbi ISE and USC CAIS) and Simon Blessenohl (USC School of Philosophy)


Vaccines for COVID-19 - Center for ImmunoEngineering - Pin Wang

The research work currently undertaken in Wang’s lab is to develop an effective vaccine against COVID-19. The Wang lab has developed several recombinant vectored vaccine platforms for cancer and infectious diseases. The team is exploring one of their platforms to create live hybrid viruses termed rVSV-COVID19, with the viral core derived from vesicular stomatitis virus (VSV) and the surface engineered with COVID-19 spike protein. This type of engineered viruses has demonstrated to be safe after testing in humans for vaccine against human immunodeficiency virus (HIV) and Ebola virus, and significant protection has been observed from Ebola virus vaccine trials. The Wang lab recognizes that no effective vaccine has been developed even for common cold-related coronaviruses, so that it is unclear at this moment as to which vaccine modalities will be effective for COVID-19. Scientific community should broadly test various vaccine designs to ensure that at least one of them will work. The research team consisted of Prof. Pin Wang and graduate students Xianhui Chen, Yun Qu, and Zachary Dunn, is working around the clock to produce rVSV-COVID19 and test it in cell culture dish for its capacity to induce immune responses. Animal studies are also planned to assess the vaccine potency once administered through either intramuscular injection or intradermal injection. One of the key parameters to measure is neutralizing antibody titer post immunization. The Wang team is also establishing experimental assays for such measurements. The team is also planning to use these assays to isolate neutralizing antibodies from COVID-19 recovered patients that can potentially be used to treat infected patients.