Artificial intelligence (AI) and a mobile phone app can be used to detect COVID-19 infection in the voice, according to researchers based in the Netherlands.
Presenting their findings on Monday at the European Respiratory Society International Congress in Barcelona, Spain, Maastricht University researchers told the congress that their AI model was accurate 89% of the time and was more accurate than lateral flow rapid antigen tests.
The researchers said that the model they developed is cheap, quick, and easy to use and could be applied in low-income countries where PCR tests are expensive and difficult to distribute.
In the study, the accuracy of lateral flow tests varied widely depending on the brand and they were less accurate at detecting COVID-19 infection in people who showed no symptoms, the researchers said.
COVID-19 infection usually affects the upper respiratory tract and vocal cords, leading to changes in a person's voice that can be used for detection.
"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," Wafaa Aljibawi, a researcher at Maastricht University's Institute of Data Science and one of the contributors to the study, said in a statement.
"Such tests can be provided at no cost and are simple to interpret," she said. "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."
The Maastricht University researchers used data from the University of Cambridge's crowd-sourced COVID-19 Sounds App that contains 893 audio samples from 4,352 healthy and non-healthy participants, 308 of whom had tested positive for COVID-19.
The app is installed on the user's mobile phone. Participants reported basic information about demographics, medical history, and smoking status, and then recorded respiratory sounds. These include coughing three times, breathing deeply through their mouth three to five times, and reading a short sentence on the screen three times.
The researchers used a voice analysis technique called Mel-spectrogram analysis, which identifies different voice features such as loudness, power, and variation over time.
"In this way we can decompose the many properties of the participants' voices," Aljbawi said. "In order 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."
They evaluated a model called Long-Short Term Memory (LSTM), which is based on neural networks that mimic the way the human brain operates and recognizes the underlying relationships in data.
It outperformed other models, having an overall accuracy of 89%, an ability to correctly detect positive cases (sensitivity) of 89%, and an ability to correctly detect negative cases (specificity) of 83%.
In a separate study presented at the congress, Henry Glyde, a University of Bristol researcher, showed that AI could be harnessed via an app to predict when patients with chronic obstructive pulmonary disease (COPD) might suffer a flareup.