No need to go anywhere to get Covid-19 test done, it wi …

New Delhi. Scientists have developed a new smartphone app that can accurately detect COVID-19 infection in people’s voices using artificial intelligence (AI). The researcher said that the AI ​​model used in the research is more accurate, faster, cheaper and easier to use than the rapid antigen test or lateral flow test.

He said the method could be used in low-income countries where PCR tests are expensive and difficult to distribute. The research was presented at the European Respiratory Society International Congress in Barcelona, ​​Spain. According to the researcher, the AI ​​model is up to 89 percent accurate, while the accuracy of the lateral flow test varies widely depending on the brand.

Furthermore, lateral flow tests are much less accurate in detecting COVID-19 infection in people who are asymptomatic, he said.

Wafa Aljabwi, a researcher at Maastricht University Netherlands, said that “these accurate results suggest that normal voice recording and fine-tune AI algorithms can achieve substantially greater accuracy in determining which patients have a COVID-19 infection.” “

Aljawabi said that “tests like these can be provided at no cost and are easy to provide information. They also work in less than a minute, helping with remote and virtual testing as well.”

The researcher said that the new test can be used at entry points in large gatherings and in rapid testing of large populations. The COVID-19 infection usually affects the upper respiratory track and vocal cords and causes changes in a person’s voice.

Aljabvi and his supervisor used data from the COVID-19 Sound app from a crowdsourced Cambridge University study, which has 893 audio samples from 4,352 healthy and non-healthy people, of whom 308 tested positive for COVID-19. The app is installed on the user’s phone. The participant is required to report some basic information about demographics, health history and smoking status and is then asked to record some respiratory sounds.

These include coughing three times, taking deep breaths through the mouth three to five times and reading a short sentence on the screen three times. The researchers used a voice analysis technology called mail-spectrogram analysis to identify characteristics of the voice with differences such as loudness, power and variation over time.

Aljabvi said that ‘in this way we can decompose many qualities of the voice of the people.’

He said that “to separate the voice of the Kovid-19 victims from those who did not have the disease. We have created several artificial intelligence models and found out which one works best in isolating COVID-19 cases. They found that a model called long-short term memory (LSTM) worked better than other models. LSTM is based on neural networks that copy the way the human brain operates and examines internal relationships in the data.

It had an overall accuracy of 89 percent. On the other hand, its ability to correctly detect positive cases or detect sensitivity properly was 89 percent and the ability or specificity of negative cases was 83 percent.

In another study, Henry Glyde, PhD student at the University of Bristol, looked at how AI can be used through an app called myCOPD to predict when people with chronic obstructive pulmonary disease (COPD) have an outbreak of their disease. could. Exacerbation of COPD can be very dangerous and may require hospitalization. Symptoms include shortness of breath, cough, and excessive mucus.