In January 2022, as Omicron pushed case counts upward, a small team at the University of the West of Scotland offered a hopeful new instrument to a world long strained by testing: an artificial intelligence that reads a chest X-ray and judges whether the lungs bear the marks of COVID-19. Its makers claim near-perfect accuracy and speed beyond the laboratory test that remains the gold standard, yet they are candid that a shadow on early infection will escape the lens, so the method is better understood as a companion to testing than its successor. Like many tools born in that crowded era, it ar
UWS Team Claims 98% Accurate AI-Assisted X-ray Test for COVID-19, But Cannot Replace RT-PCR Yet
It cannot fully replace RT-PCR, but may help where PCR tests are scarce
Mark
So what exactly did the Scottish team build?
Mimi
An AI system that reads chest X-rays. It compares a scan against about 3,000 images of COVID patients, healthy people, and people with viral pneumonia, then makes a call.
Mark
And the headline number is 98 per cent accuracy?
Luke
Whose number, though? It's the team's own claim. The report doesn't say anyone outside UWS has verified it.
Mimi
That's fair. It's reported as the researchers' figure, not an independent result.
Mark
Why include viral pneumonia in the image set?
Mimi
The source doesn't explain it. It simply lists it as one of the three groups. I'd guess it helps separate COVID from similar-looking illness, but that's inference, not reporting.
Luke
Right, and I'd keep it labelled as a guess.
Mark
Can it replace RT-PCR?
Mimi
Ramzan says no. Symptoms don't show on X-rays early in infection, so it would miss those cases.
Luke
That's a significant gap, since early detection is much of the point of testing.
Mark
So where does it help?
Mimi
Where PCR tests are scarce. Ramzan also says results come faster than PCR's roughly two hours.
Luke
We don't know how much faster, or how it would be validated in real clinics. Those answers aren't in the report.
Il Polso
- Omicron is driving a rise in cases, and the demand for fast, reliable diagnosis is pressing harder than the supply of tests can meet.
- A Scottish team claims an AI can read chest X-rays to detect COVID-19 with 98 per cent accuracy, faster than a two-hour RT-PCR result.
- The tool's own creators admit X-rays miss infections in early stages, so it cannot replace RT-PCR and must be weighed as a supplement.
- The accuracy figure comes from the researchers alone, and it awaits independent validation before anyone can lean on it fully.
In January 2022, as Omicron pushed case counts upward, a small team at the University of the West of Scotland offered a hopeful new instrument to a world long strained by testing: an artificial intelligence that reads a chest X-ray and judges whether the lungs bear the marks of COVID-19. Its makers claim near-perfect accuracy and speed beyond the laboratory test that remains the gold standard, yet they are candid that a shadow on early infection will escape the lens, so the method is better understood as a companion to testing than its successor. Like many tools born in that crowded era, it arrives as a promise still waiting for the wider world to confirm it.
Scientists at the University of the West of Scotland say they have built a way to diagnose COVID-19 from an X-ray, letting artificial intelligence read the image in place of waiting on a laboratory result. Their claim of 98 per cent accuracy, and results faster than the standard test, was reported by Firstpost in January 2022, at a moment when the Omicron variant was lifting case numbers again.
The method compares a patient's scan against a database of roughly 3,000 images drawn from people with COVID-19, healthy individuals, and patients with viral pneumonia. An algorithm analyses the visual patterns and reaches a diagnosis. Professor Naeem Ramzan led the three-person team, joined by Gabriel Okolo and Dr Stamos Katsigiannis.
Ramzan framed the work as an answer to a long-standing call for a quick, reliable detection tool, a need sharpened by Omicron. Many countries, he noted, cannot run large testing programmes because their diagnostic tools are limited, while this technique relies on technology that is easy to access.
He was equally plain about its limits. Symptoms do not show on X-rays in the earliest stages of infection, so the method cannot fully replace RT-PCR, the gold standard that reveals the virus present at the time of testing. Still, he believes it could help slow the spread, particularly where PCR tests are scarce. Professor Milan Radosavljevic, the university's Vice-Principal for Research, Innovation and Engagement, and also part of the team, called the work potentially game-changing.
The figure of 98 per cent comes from the researchers themselves, and the report does not say it has been independently verified. Until wider validation arrives, the tool looks less like a replacement for testing and more like a support for places short of PCR capacity.
Scientists at the University of the West of Scotland say they have built a way to diagnose COVID-19 from an X-ray, using artificial intelligence to read the image. The team claims the method is 98 per cent accurate and returns results faster than the standard laboratory test. Firstpost reported the claim in January 2022, as the Omicron variant was driving a rise in cases.
The method works by comparing a patient's X-ray scan against a database of around 3,000 images. Those images come from three groups: people with COVID-19, healthy individuals, and patients with viral pneumonia. An AI process then uses an algorithm to analyse the visual imagery and reach a diagnosis.
Professor Naeem Ramzan led the three-person team. He directs the Affective and Human Computing for SMART Environments Research Centre at UWS. Gabriel Okolo and Dr Stamos Katsigiannis worked with him on the project.
Ramzan framed the work around a long-standing need. He said there has been a call for a quick, reliable tool to detect the virus, and that the need has grown with Omicron. In his words, several countries cannot carry out large numbers of tests because diagnostic tools are limited, whereas this technique relies on technology that is easy to access.
He was also direct about the limits. COVID symptoms do not show up on X-rays in the early stages of infection, so the method cannot fully replace RT-PCR testing. Still, he said, it could play an important part in slowing the virus's spread, especially where PCR tests are not readily available. He added that the new method is quicker than an RT-PCR test, which typically takes around two hours.
Professor Milan Radosavljevic, UWS Vice-Principal of Research, Innovation and Engagement, who is also part of the team, called the work potentially game-changing. He described it as another example of purposeful, impactful work at the university during the pandemic.
For context, the report lays out the tools already in use. The US Centers for Disease Control and Prevention says COVID-19 tests detect either SARS-CoV-2, the virus itself, or the antibodies the body makes after infection or vaccination. RT-PCR detects genetic material from a specific organism and shows whether someone has the virus at the time of the test. It is regarded as the gold standard. The antigen test looks for proteins on the virus's surface and typically takes 15 to 30 minutes, though it is not considered as accurate.
The 98 per cent figure comes from the researchers themselves, and the report does not say it has been independently verified. Because of the early-infection blind spot, the tool looks more like a supplement for places short of PCR capacity than a replacement. Wider validation is the thing to watch.
Citazioni salienti
It can still play an important role in curtailing the virus's spread, especially when PCR tests are not readily available.— Professor Naeem Ramzan, UWS
This is potentially game-changing research.— Professor Milan Radosavljevic, UWS
Vuoi la storia completa?
Leggi l'originale su Firstpost ↗
Technology