A novel artificial intelligence (AI) model has been developed that can detect COVID-19 by listening to someone’s voice through a smartphone application. The groundbreaking findings were presented at the European Respiratory Society International Congress on Monday in Barcelona, Spain.
According to a report by News Medical, the AI model employed in this study is suitable for application in low-income nations where PCR tests are expensive or challenging to distribute.
Although the accuracy of lateral flow testing varies greatly depending on the brand, the AI model called the LSTM (Long-Short Term Memory) has 89% accuracy of the time, as said by Wafaa Aljbawi, a researcher at the Institute of Data Science at Maastricht University in the Netherlands. Positive cases were detected with 89 percent accuracy, and negative cases were identified with 83 percent accuracy.
In those with COVID-19 infection, lateral flow tests were significantly less likely to detect those without symptoms.
A technique called Mel-spectrogram analysis was used to identify specific voice features suggesting possible contraction of Covid-19, including loudness, power, and variation over time.
Using basic voice recordings and customized AI algorithms has demonstrated promising results in identifying COVID-19-infected patients with high precision. Aljbawi says that such tests can be delivered for free and are straightforward to interpret. Moreover, Aljbawi says they enable remote, virtual testing and have a turnaround time of less than a minute.
More often than not, the upper respiratory tract and vocal cords are both affected by COVID-19, altering a person’s voice. As part of the study, Visara Urovi, from the Institute of Data Science, and Dr. Sami Simons, a pulmonologist at Maastricht University Medical Center, investigated whether artificial intelligence could be used to analyze voices and identify COVID-19.
Users install the app on their phones. After reporting some basic demographics, medical history, and smoking status, participants are asked to record a few respiratory sounds. These include breathing deeply through the mouth 3 to 5 times, coughing 3 times, and reading a short sentence on the screen 3 times.
A crowdsourced COVID-19 Sounds App from the University of Cambridge was used by the research team to collect audio samples from 4,352 healthy and unhealthy subjects. Out of these subjects, 308 received COVID-19 positive test results, based on information collected from the COVID-19 Sounds App.
