Novel Signal Processing Technology to Significantly Scale Up COVID-19 Test Throughput

COVID-19 pandemic outbreak has significantly harmed or threatened the health and lives of millions or even billions of people. Mass testing of COVID-19 viruses and antibodies are essential for disease diagnosis, virus spread confinement, contact tracing, and determining right conditions/right candidates to return to workforce. However, the throughputs of current testing technologies for COVID-19 viruses and antibodies are often quite limited, and there are not enough reagents to perform needed mass testing.

Together with graduate student Jirong Yi,  Electrical and Computer Engineering Professors Weiyu Xu, Raghu Mudumbai and Xiaodong Wu have originated a new high-throughput and low-reagent-consumption method of using compressed sensing, a novel mathematical and signal processing idea, to significantly speed up testing COVID-19 virus and antibody.  Preliminary results show that this method can potentially provide ten or even more folds of speedup compared with current testing technologies. For example, this new virus testing method can potentially increase COVID-19 test capacity of a lab from 3,000 tests per day to 30,000 tests per day.  The basic idea is to test the quantity/presence of virus/antibody in mathematically-well-designed mixed samples (for example, using real-time quantitative polymerase chain reaction machines), and use signal processing algorithms to infer the infectious status of each person.  For example, consider testing three persons for possible virus infections. The authors' method would mix the swab samples of Person 1 and Person 2 together, do the first real-time PCR test for this mixed sample; and then pool the the swabs samples of Person 2 and Person 3 together, do the second PCR test for this mixed sample. If the first test reveals 0 virus, and the second test reveal 8 viruses (hypothetically), then immediately we are able to infer Person 1 and Person 2 test NEGATIVE, and Person 3 tests POSITIVE with 8 viruses.  For this example, only 2 tests, instead of 3 tests, are able to give the virus infection status of three persons.  More sophisticated sample mixing designs such as through sparse bipartite expander graph can increase the test throughput even more significantly.

To the best of the authors' knowledge, their work might be the first to propose and design compressed sensing method for virus/antibody detection in general, and for COVID-19 detection in particular. A related classical method ``group testing'' was invented by American applied mathematician Robert Dorfman in World War II to detect soldiers with syphilis. Compared with group testing, the authors' compressed sensing method  uses fewer tests, offers larger scale-up of test throughput, and can give quantitative rather than qualitative test results. Recently there have been preliminary experiments performed by other groups of researchers from India IIT and Harvard University validating the compressed sensing idea for virus testing, and have cited the authors' work in their preprints on   Read more about the work on arXiv.