The COVID-19 pandemic has resulted in more than 200 million infections, and more than 4 million casualties. Although the case-to--fatality ratio (CFR) is less than 1/5th of that in the influenza pandemic in 1928, in the current age of science and technology, these figures are worrying. The pathogenesis of COVID-19 is increasingly suggesting impairments in the respiratory system. In this light, it is natural to ask - Can sound samples serve as acoustic biomarkers of COVID-19? If yes, an acoustics based COVID-19 diagnosis can provide a fast, contactless and inexpensive testing scheme, with potential to supplement the existing molecular testing methods, such as RT-PCR and RAT. The DiCOVA Challenge Series is an exploration of ideas to find answers to this question.
Considering the immediate societal relevance of a technology driven point-of-care-test (POCT) for COVID-19, the DiCOVA Challenge has three aims. Release a curated dataset of sound samples (breathing, cough, and speech) drawn from individuals with and without COVID-19 during the time of recording. Invite researchers from around the globe to search for acoustic biomarkers in this dataset. Evaluate the findings of each group using a blind test set, and present a competitive leaderboard with global participation.
Building on the overwhelming response to the first DiCOVA Challenge, we are excited to launch this Second DiCOVA Challenge with the following updates. Larger dataset: Over the past few months we have created an audio dataset with an improved ratio of COVID-19-to-non-COVID subjects. Multi-modality: In addition to the cough sound samples used in the first DiCOVA challenge, recent research indicates the COVID-19 acoustic signature can be also extracted from breathing and speech sound samples. The second DiCOVA challenge will feature three sound categories (cough, speech and breathing) and four tracks (three individual categories and one fusion). A leaderboard will be set to track the performance as done in the previous challenge. Enhanced Pre-training: Several open source and publicly available cough datasets have been published recently. The challenge participants will be encouraged to use these datasets to pre-train their models for improved performance. We are hopeful, this will open up new horizons for focussed and timely effort from researchers on a topic at the interface of acoustics, signal processing, and healthcare.
We look forward to your participation!
Prior to this challenge, the first DiCOVA Challenge was launched on Feb 04, 2021. It focussed on COVID-19 detection using only cough sounds. The challenge received an enthusiatic response from the research community across industry and acadmia. The challenge closed on Mar 23, 2021. A summary of results is provided here and here.
The challenge features four tracks. The first three tracks focus on single sound categories, and the fourth track is a fusion track. The task is to design a binary classifier for COVID-19 detection.