A study finds that artificial intelligence (AI) can detect Covid-19 in people's voices via a phone app. Researchers say it is also more accurate than lateral flow and rapid antigen tests.
Scientists say the AI model used in the app is 89% accurate and is also considerably better than regular tests at detecting Covid in people who show no symptoms.
It provides results in less than a minute, is easier to use, and is also cheaper, which makes it particularly practical in low-income countries where PCR tests are expensive and difficult to distribute.
"These promising results suggest that simple voice recordings and fine-tuned AI algorithms can potentially achieve high precision in determining which patients have Covid-19 infection," said Wafaa Aljbawi, a researcher at the Institute of Data Science at the Netherlands-based Maastricht University.
"Such tests can be provided at no cost and are simple to interpret. Moreover, they enable remote, virtual testing and have a turnaround time of less than a minute. They could be used, for example, at the entry points for large gatherings, enabling rapid screening of the population," Aljbawi said.
The Covid-19 infection usually affects the upper respiratory tract, as well as vocal cords, which causes a person’s voice to change.
"Digital health using AI models presents an exciting opportunity and is likely to impact future health care," said Professor Chris Brightling, who chairs the European Respiratory Society (ERS) Science Council and was not involved in the research.
The results of the research were presented at the ERS International Congress in Barcelona on Monday.
Model based on neural networks
Aljbawi and her team used data from Cambridge University's crowd-sourcing Covid-19 Sounds App. It included 893 audio samples from 4,352 healthy and non-healthy people.
Upon installing the app, users are asked to disclose their demographic background, medical history, and smoking status. They will then need to record some respiratory sounds, such as coughing, breathing deeply through the mouth, and reading a short sentence several times each.
The research used a voice analysis technique called Mel-spectrogram analysis. It identifies different voice features, such as loudness or power.
"In this way, we can decompose the many properties of the participants' voices," Aljbawi said. "To distinguish the voice of Covid-19 patients from those who did not have the disease, we built different artificial intelligence models and evaluated which one worked best at classifying the Covid-19 cases."
The research team found that a model based on neural networks, called Long-Short Term Memory (LSTM), outperformed others. Neural networks mimic the way human brains operate and recognize the underlying relationships in data.
"The lateral flow test has a sensitivity of only 56%, but a higher specificity rate of 99.5%. This is important as it signifies that the lateral flow test is misclassifying infected people as Covid-19 negative more often than our test," Aljabawi said.
"In other words, with the AI LSTM model, we could miss 11 out 100 cases who would go on to spread the infection, while the lateral flow test would miss 44 out of 100 cases," she noted.
The AI model is also being used to develop an app that can predict exacerbations in chronic obstructive pulmonary disease.
Your email address will not be published. Required fields are markedmarked