Artificial intelligence (AI) can be used to detect COVID-19 infection in people’s voices by means of a mobile phone app, according to research to be presented on Monday at the European Respiratory Society International Congress inside Barcelona, Spain.
The AI model used in this research is more accurate than lateral flow/rapid antigen tests and is cheap, quick and easy to use, which means it can be used in low-income countries where PCR tests are expensive and/or difficult to distribute.
Ms Wafaa Aljbawi, a researcher in the Institute of Data Science, Maastricht University, The Netherlands, told the congress that the AI model was precise 89% of the time, whereas the particular accuracy associated with lateral flow tests varied widely depending on the brand. Also, lateral circulation tests were considerably less accurate at detecting COVID infection within people who showed no symptoms.
These promising results suggest that simple voice recordings plus fine-tuned AI algorithms can potentially achieve high precision in determining which patients have COVID-19 contamination. Such tests can be provided in no cost and are simple to interpret. Moreover, they enable remote, virtual testing and have a turnaround time of much less than a minute. They could become used, for example, on the entry points for large gatherings, enabling rapid screening of the particular population. ”
Wafaa Aljbawi, Researcher, Company of Information Science, Maastricht University
COVID-19 illness usually affects the upper respiratory track and vocal cords, leading to changes in a person’s voice. Microsoft Aljbawi plus her supervisors, Dr Sami Simons, pulmonologist at Maastricht University Medical Centre, and Dr Visara Urovi, also from the Start of Data Science, decided to investigate if it was possible to utilize AI to analyze sounds in order to detect COVID-19.
These people used data from your University of Cambridge’s crowd-sourcing COVID-19 Sounds App that contains 893 audio samples from 4, 352 healthy and non-healthy participants, 308 of whom had tested positive with regard to COVID-19. The particular app is installed on the user’s mobile phone, the particular participants report some basic information about demographics, medical history plus smoking status, and then are asked to record some respiratory system sounds. These include coughing three times, breathing deeply through their mouth three to five times, and reading the short sentence on the screen 3 times.
The researchers used a voice analysis technique called Mel-spectrogram analysis, which usually 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, inch said Ms Aljbawi. “In order to distinguish the tone of voice of COVID-19 patients from those that did not possess the disease, we built different artificial intelligence models and evaluated which one worked best at classifying the COVID-19 cases. inches
They found that one model called Long-Short Term Memory (LSTM) out-performed the other models. LSTM will be based on neural networks, which mimic the way the human brain operates and recognizes the underlying relationships in data. It works with sequences, which makes it suitable regarding modeling signals collected over time, such as from the particular voice, because of its ability to store data within its memory.
Its overall accuracy was 89%, the ability in order to correctly identify positive cases (the true positive rate or “sensitivity”) was 89%, and its ability to correctly identify negative cases (the true negative rate or “specificity”) was 83%.
“These results show a significant improvement in the accuracy of diagnosing COVID-19 compared to state-of-the-art assessments such while the horizontal flow test, ” said Ms Aljbawi. “The spectrum of ankle flow check has a sensitivity associated with only 56%, but the higher specificity rate of 99. 5%. This is usually important as it signifies that the lateral movement test is definitely misclassifying infected people as COVID-19 negative more often than our test. In other words, with the AI LSTM design, we could miss 11 out 100 instances who would go on to spread the infection, while the lateral stream test would miss 44 out of 100 cases.
“The high specificity of the lateral flow test means that only one within 100 people would end up being wrongly informed they were COVID-19 positive when, in fact , they were not infected, while the LSTM check would wrongly diagnose 17 in one hundred non-infected individuals as positive. However, since this test is virtually free, it is possible to invite people for PCR tests if the LSTM tests show they are positive. ”
The researchers say that their results need to be validated with big numbers. Since the start of this project, 53, 449 audio samples through 36, 116 participants have now been collected and can be used in order to improve plus validate the particular accuracy from the model. They are also carrying out there further analysis to understand which parameters in the voice are influencing the AI model.
In a second study, Mr Henry Glyde, a PhD student in the faculty associated with engineering from the College of Bristol, showed that will AI could be harnessed via an app called myCOPD to predict when patients with chronic obstructive pulmonary disease (COPD) might suffer a flare-up of their own disease, sometimes called acute exacerbation. COPD exacerbations may be very serious and are associated with increased risk of hospitalization. Symptoms include shortness of breath, hacking and coughing and producing more phlegm (mucus).
“Acute exacerbations associated with COPD have got poor outcomes. We know that early identification and treatment of exacerbations can improve these outcomes plus so all of us wanted to determine the particular predictive capability of a widely used COPD app, inch he stated.
The myCOPD app is a cloud-based interactive app, developed by patients and clinicians and is available in order to use in the UK’s National Health Service. It was established in 2016 and, so far, has over 15, 000 COPD patients using it to help them manage their illness.
The experts collected 45, 636 records for 183 patients between August 2017 and December 2021. Of these, forty five, 007 had been records associated with stable condition and 629 were exacerbations. Exacerbation predictions were generated one to eight days before the self-reported exacerbation event. Mr Glyde plus colleagues utilized these information to train AI models upon 70% of the data and test it on 30%.
The patients were “high engagers”, who had been using the particular app weekly over months or even years to record their symptoms and other health information, report medication, set reminders, and have access in order to up-to-date wellness and lifestyle information. Doctors can assess the information via a clinician dashboard, enabling them to provide oversight, co-management and remote control monitoring.
“The most recent AI model we developed has a sensitivity of 32% and a specificity of 95%. This means that the model can be very good at telling individuals when they are not regarding to experience an excitement, which may help all of them to avoid unnecessary treatment. It is certainly less good at telling them when they are about to encounter one. Improving this will be the particular focus of the next phase of our own research, inches said Mister Glyde.
Speaking before the our elected representatives, Dr James Dodd, Associate Professor inside respiratory medicine at the University or college of Bristol and project lead, mentioned: “To the knowledge, this study is the first of its kind to design real world data from COPD sufferers, extracted through a widely deployed therapeutic app. As a result, exacerbation predictive models produced from this particular study have got the potential to be deployed to thousands more COPD patients after further safety and efficacy screening. It would empower patients in order to have a lot more autonomy plus control more than their health. This is furthermore a substantial benefit for their doctors as such a system would likely reduce patient reliance on primary care. Within addition, better-managed exacerbations could prevent hospitalization and alleviate the burden on the healthcare system. Further study is required into patient engagement to determine what level of accuracy is acceptable and how an exacerbation alert system would work in practice. The introduction of sensing technologies may further enhance monitoring and improve the predictive performance of models. ”
One of the limitations from the study is the small number of frequent users of the app. The current model requires a patient to input the COPD assessment test score, fill out their medication diary and then statement they are usually having an exacerbation accurately days later. Usually, just patients who are highly engaged using the app, utilizing it daily or even weekly, can provide the amount of data needed intended for the AI modeling. In addition, because there are significantly more days the users are stable than when they are having a good exacerbation, there is a significant imbalance between the excitement and non-exacerbation data available. This outcomes in even further difficulty in the models properly predicting events after training on this imbalanced information.
“A current partnership among patients, clinicians and carers to set research priorities within COPD found out that this highest-rated question has been how to identify better ways to prevent exacerbations. We have focused on this question, and we will become working closely with individuals to design plus implement the particular system, ” concluded Mr Glyde.
Chair of the ERS Science Council, Professor Chris Brightling, may be the Nationwide Institute to get Health and Care Research (NIHR) Senior Investigator in the College or university of Leicester, UK, and was not involved with the research. He commented: “These two studies display the potential of synthetic intelligence and apps upon mobile phones plus other digital devices to make a difference in how diseases are managed. Having more data available for training these types of artificial cleverness models, including appropriate control groups, as well as validation within multiple studies, will improve their precision and reliability. Digital wellness using AI models presents an exciting opportunity and it is likely to impact future health care. inch