The Pandemic has literally and figuratively sucked the life out of the earth for the past year and a half. People worldwide have been facing challenges, and every second of this prolonged suffering proves fatal for someone somewhere. Science has been trying to cope with it, and we have made significant strides in the form of tests and vaccines. An addition to the science column is the new AI tech which detects even asymptomatic Covid positivity with just a cough.
Australian scientists have developed an Artificial Intelligence-based method that can hear the effects of Covid-19 through the sound of someone coughing forcefully, even when folks around you are asymptomatic. This is an advanced tech that can pave the way for observing infectious disease via a diagnostic smartphone app.
With this Pandemic growing globally, many crowdsourcing platforms have been developed to huddle respiratory sound audios from both healthy and Covid-19 positive groups for research purposes. This research conducted by RMIT University is proven because they have accessed datasets from two of these platforms, Covid-19 Sounds App and COSWARA, to educate the algorithm using contrastive self-supervised understanding, a method by which the whole system works unassisted to encode what gives rise to two things that are similar or different.
With additional development, their algorithm could strength a diagnostic mobile phone app, said lead author Hao Xue, Research Fellowship in RMIT’s School of Computing Technologies branch.
“We’ve overcome some significant challenges in the development of a more reliable, easier to use, accessible and contactless preliminary diagnosis tool for Covid-19,” said Xue, Research Fellow in RMIT’s School of Computing Technologies branch. This could prove to be a substantial advantage in hindering the spread of the virus by those who retain no apparent symptoms. “A mobile app that can give you a heads up during community outbreaks or prompt you to seek a Covid test, that’s the kind of clever tool we desire to oversee this pandemic generously,” Xue added.
Xue said this method they developed could also be broadened for other respiratory diseases like tuberculosis, etc. While this isn’t the first Covid cough sequence algorithm to be developed, the RMIT model, like said, outperforms its predecessors.
According to co-author Professor Flora Salim, prior endeavors to evolve this type of technology, like those at MIT and Cambridge, banked vast amounts of meticulously interpreted data to train the AI system. But this annotation of respiratory sounds compels specific knowledge from experts, making it costly and time-consuming, and complicates handling sensitive health information,” Flora Salim added.
Moreover, cough specimens from one hospital or one area to train the algorithm also restricts its performance outside that pertained environment. What’s most fascinating about our work is we have withstood this problem by developing a method to train the algorithm using unlabeled cough sound data. This can be acquired relatively quickly and at a larger scale from different countries, genders and ages,” Salim pointed out.
Further detailed information about this research is published from a syndicated feed. But summing up, this research does stand alone and proves more helpful and easy to use in these tough times. However, the development is going on, and we need to wait for further details.