AI needs a Cough to detect Asymptomatic COVID-19 cases
MIT has developed an AI that can say if you are having COVID-19 from your cough.
AI has been at forefront of battling against deadly coronavirus caused COVID-19. While it is easy to detect symptomatic cases via nasal swabs, asymptomatic ones are challenging. Fortunately, AI can help to resolve this too, apart from assisting in drug discovery, chatbots, thermal screening, and others. Recently, researchers at the Massachusetts Institute of Technology (MIT) developed an AI algorithm using neural network that accurately spots those infected with coronavirus based on their cough. And this algorithm helps identify asymptomatic COVID-19 patients from healthy individuals by just a forced cough recording and shows the results on a smartphone app. In a paper published in the IEEE Journal of Engineering in Medicine and Biology, the team describes that AI can pick up the cough that may not be decipherable even to the human ear. The AI has 98.5 percent accuracy and was trained on thousands of samples of coughs and spoken words. MIT scientist Brian Subirana, who co-authored the paper, said: “The way you produce sound changes when you have COVID, even if you’re asymptomatic.”
Over 70,000 samples have been collected so far, and roughly 2,500 were submitted by individuals confirmed to have COVID-19. These recordings were submitted by people voluntarily through web browsers and devices such as cellphones and laptops. The researchers used 4,000 of these samples to train the AI model. The remaining 1,000 recordings were then fed into the model to see if it could accurately discern coughs from COVID patients versus healthy individuals. The researchers were able to pick up patterns in the four biomarkers — vocal cord strength, sentiment, lung and respiratory performance, and muscular degradation — that are specific to COVID-19.
The AI, speech processing framework, controls the acoustic biomarker feature extractors to assess COVID-19 from cough recordings. It provides an individualized patient saliency map to monitor patients in real-time. Plus, it is non-invasive and cost-effective.
Apart from Brian, MIT’s team comprised of Jordi Laguarta and Ferran Hueto from MIT’s Auto-ID Laboratory. The team is working on incorporating the model into a user-friendly app, after receiving approval from the Food and Drug Administration (FDA), and adopted on a large scale that could potentially be a free, convenient, non-invasive pre-screening tool. This can help to identify people who are likely to be asymptomatic for COVID-19, the researchers said. A user could log in daily, cough into their phone, and instantly get information on whether they might be infected and therefore should confirm with a formal test. MIT is now working with a number of hospitals that will provide additional recordings to improve the model further. This research was supported, in part, by Takeda Pharmaceutic.
Prior to researching on COVID-19, Brian and his team were working on using an AI algorithm that could detect Alzheimer’s disease. The team stated that generally, people associate Alzheimer’s with its degenerative neurological effects. Though, it also causes degradation of the neuromuscular system—especially the vocal cords. This, in theory, makes it possible to hear changes in the vocalizations of someone with the disease.
The AI model was a combination of three neural networks. The Alzheimer’s research was repurposed for COVID-19 involved a neural network known as ResNet50. It was trained on a thousand hours of human speech. The second neural network was trained to distinguish between different emotional states evident in speech. The third neural network was trained then on a database of coughs to discern changes in lung and respiratory performance. When the three models were combined, a layer of noise was used to filter out stronger coughs from weaker ones.
Based on the testing, the researchers said their tool discriminated COVID-19 positive participants with 97.1% accuracy, 98.5% sensitivity and 94.2% specificity. Of particular note, the model performed at 100% accuracy when detecting coughs from asymptomatic positive cases.
The paper concludes that practical use cases of this AI algorithm can be for daily screening of students, workers, and public as schools, jobs, and transport reopen, or for pool testing to alert of outbreaks in groups quickly.”
Brian says, “The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant”. “Pandemics could be a thing of the past if pre-screening tools are always on in the background and constantly improved,” he added.