AI for Pooled Testing of COVID-19 Samples

The COVID-19 pandemic has adversely affected millions all over the world. Efficient and effective testing of individuals for COVID-19, via modalities such as reverse transcription polymerase chain reaction (RT-PCR) is a crucial factor in combating this menace. Given the widespread scarcity of testing resources including testing kits, reagents, skilled manpower and available time, pooled testing has been advocated as a method of speed-up. Pooling involves mixing together small portions of ‘samples’ of different individuals, followed by testing the pools instead of the individual samples. It has been observed that a much smaller number of pools, as compared to the number of samples, is sufficient to allow for accurate prediction of the health status of the constituent samples, under the common and reasonable assumption that only a small number of the samples were infected. Artificial intelligence (AI) has emerged as a key tool in improving the prediction accuracy as well as efficiency of pooled testing. Such algorithmic tools are often studied within the frameworks of group testing and compressed sensing. In this chapter, we present algorithmic tools for pooled testing and recovery, giving a broad description of the use of AI for pooled testing in the context of COVID-19.