• Sat. Oct 1st, 2022

Mobile app detects Covid-19 from the sound of someone’s voice

ByCindy J. Daddario

Sep 5, 2022

Scientists at Maastricht University have developed a mobile phone app that can detect Covid-19 infections “more accurately than lateral flow tests”.

Artificial intelligence (AI) can be used to detect Covid-19 infection in people’s voices through a mobile phone app, according to a study to be presented at the European Respiratory International Congress on Monday Society in Barcelona, ​​Spain.

One of the main symptoms of Covid-19 is inflammation of the upper airways and vocal cords, which usually leads to changes in the patient’s voice. Therefore, scientists from Maastricht University decided to study whether these symptoms could be used as an accurate method to diagnose the disease, especially in low-income countries. where PCR tests are expensive or difficult to distribute.

With an accuracy rate of 89%, the AI ​​model was able to perform more accurate Covid-19 diagnoses than rapid antigen tests, according to Wafaa Aljbawi, researcher at the Institute for Data Science at Maastricht University.

“These promising results suggest that simple voice recordings and fine-tuned AI algorithms can potentially achieve high accuracy in determining which patients are infected with Covid-19,” she said. “Such tests can be provided free of charge and are simple to interpret. Additionally, they allow remote virtual testing and have a turnaround time of less than a minute. They could be used, for example, at the entry points of large gatherings, allowing rapid screening of the population.

To train the algorithm, Aljbawi and his team used data from the University of Cambridge’s Covid crowd-sourcing-19 sounds app, who contains 893 audio samples from 4,352 healthy and unhealthy participants, 308 of whom had tested positive for Covid-19.

The application they have developed can be installed on the user’s mobile phone. To use it, participants report some basic information about demographics, medical history, and smoking status, then are asked to record breath sounds. These include coughing three times, taking a deep breath through your 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 characteristics such as loudness, loudness and variation over time, to make the diagnosis.

“In order to distinguish the voice of Covid-19 patients from those without the disease, we built different artificial intelligence models and assessed which worked best for classifying Covid-19 cases,” said Aljbawi.

The model that performed best was the one known as Long-Short Term Memory (LSTM). LSTM is based on neural networks, which mimic the functioning of the human brain and recognize underlying relationships in data. It works with sequences, which makes it suitable for modeling signals collected over time, such as voice, due to its ability to store data in memory.

The app’s overall accuracy was 89%, as was its ability to correctly detect positive cases (the true positive rate or “sensitivity”). Its ability to correctly identify negative cases (the true negative rate or “specificity”) was 83%.

In contrast, lateral flow tests have a sensitivity of 56%, but a higher specificity rate of 99.5%. This would mean that the tests more often classify infected people as Covid-19 negative than the scientists’ AI.

“These results show a significant improvement in the diagnostic accuracy of Covid-19 compared to state-of-the-art tests such as the lateral flow test,” Aljbawi said. “In other words, with the AI ​​LSTM model, we might miss 11 out of 100 cases that continue to spread infection, while the lateral flow test would miss 44 out of 100 cases.”

The researchers stressed that their results need to be validated with large numbers. To this end, they collected 53,449 audio samples from 36,116 participants, which they plan to use to improve and validate the accuracy of the model. They also perform deeper analysis to understand which voice parameters influence the AI ​​model.

Sign up for the E&T News email to get great stories like this delivered to your inbox every day.