Logiciel IA pour repérer le virus

Un logiciel d'IA pour repérer plus vite le coronavirus

Jeudi, 09/04/2020

Depuis quelques semaines, les médecins de l'hôpital de Zhongnan, à Wuhan, en Chine, peuvent recourir à un nouveau logiciel pour

détecter le coronavirus. Reposant sur la vision par ordinateur, ce système a été développé par la start-up Infervision. Fondée en 2015

et basée à Pékin, la jeune pousse est membre de l'incubateur de l'américain Nvidia.

Après s'être entraîné sur plus de 2000 clichés d'imagerie médicale de patients atteints du coronavirus, ce logiciel s'est montré

capable de détecter plus rapidement qu'un spécialiste les signes cliniques d'une pneumonie, pathologie associée au Covid-19, via des images de scanners thoraciques.

Cet outil doit permettre de prioriser la prise en charge des patients présentant cette pathologie pour effectuer des tests complémentaires et confirmer le diagnostic. Ce logiciel, bien qu'il n'ait pas encore reçu les autorisations officielles de mise sur le marché, a été déployé dans 34 hôpitaux en Chine et utilisé pour plus de 32 000 cas pour l'instant. Il serait actuellement en phase de test dans des hôpitaux européens et américains pour une future utilisation.

Une récente étude de l'Université de Séoul a montré l'importance de l'IA appliquée à l'imagerie médicale pour massivement détecter les signes cliniques du coronavirus et soulager le lourd travail des radiologues et des cliniciens submergés par la multiplication de cas.

Article rédigé par Georges Simmonds pour RT Flash

Interesting Engineering

Doctors in China Are Using AI to Screen COVID-19 Patients

Doctors in China have a new weapon in their arsenal against COVID-19, an AI-based screening software.

By Christopher McFadden / March 02, 2020 / inferVISION

Doctors in China have been given a new powerful tool to help them quickly diagnose potential coronavirus sufferers. Called inferVISION, this AI-based software can quickly highlight potential problem cases in record time.

Physicians in China are using AI to help detect coronavirus

A team of physicians in Wuhan, China, at the Zhongnan Hospital are using GPU-accelerated software to detect the visual signs of COVID-19. This AI-based software relies on NVIDIA GPUs for both training and inference and is alleviated the pressure on overworked staff to screen patients for the virus.

The software is greatly helping medical staff to prioritize those who are likely to have contracted the virus.

The software has been developed by a Beijing-based startup called inferVISION. This company is a member of NVIDIA's AI incubator Inception.

inferVISION is helping doctors quickly screen patients, Source: Manjurul/iStock

"inferVISION is a leading global high-tech enterprise in medical artificial intelligence. Our corporate goal is to empower doctors with higher efficiency and benefit patients with better diagnosis, outcome and lower cost.

00:20 / 00:20TOP ARTICLES2/5British Town Uses Oregon’s ‘Exploding Whale’ Fiascoto Teach Physical Distancing

inferVISION is using AI and deep learning technologies to develop multiple platforms, including an AI applications management platform, an AI data-mining research platform, and several AI clinical application platforms, as well as medical AI systems for quality control, health management, and scientific research.

inferVISION provides advanced, intelligent, and systematic services to payors, providers, and patients, to truly realize "Advancing Technology, Inspiring Healthcare."

The software relies heavily on NVIDIA's Clara SDKs, which is NVIDIA's AI healthcare application framework for AI-powered Medical Imaging.

inferVISION is able to identify typical signs or partial signs of COVID-19 in suspected patients. In order to do this, the software looks out for signs of pneumonia that can be caused by the virus.

Controversial World's First Monkey-Pig Hybrids Born in China

The software had originally been developed for detecting signs of cancer in lung

CTs. Additionally, the software's models were already set up for detecting

pneumonia prior to the outbreak of coronavirus and only needed some fine-tuning

to be adapted.

The AI behind the software was trained using more than 2,000 CT images of some

of the first confirmed coronavirus patients in China.

The AI-based software is proving very effective

NVIDIA's AI-based software has some fantastic advantages over its human counterparts. Foremost among them is its speed.

More conventional methods can take a considerable amount of time, especially with the current enormous demand for this kind of screening. With the software in their arsenal, doctors are able to rapidly determine the best treatment options for patients suffering from the virus.

Thus far the new end-to-end system has been deployed to at least 34 hospitals in China. It has also been used the review well over 32,000 cases of COVID-19 infected patients.

The software has been trained using hundreds of thousands of lung images collected by hospitals all across China. In addition to being used in China, inferVISION’s software is currently in evaluation by clinics in Europe and the United States.

“inferVISION is fully committed to support the medical professionals and stand with the patients in controlling the 2019-nCoV outbreak,” the company stated in a LinkedIn post.

This is not the first AI software being used to fight the COVID-19 outbreak

Interestingly this isn't the first time the power of AI has been employed to help fight the coronavirus. Machine-learning programs are being used to analyze websites, news reports and social media posts for signs of the symptoms like fever and breathing problems.

An international team of researchers, including John Brownstein from the Harvard Medical School, are collaborating to mine the internet to get a heads up on potential problem areas around the world. The program is specially designed to look for mentions of specific symptoms, like respiratory problems and fever, from geographical areas where doctors have reported potential cases.

By using natural language processing, the AI is able to disseminate between posts discussing the news and those posts that appear to be someone complaining about how they are feeling.

Another company called BlueDot has also used a similar technique albeit without social media sources. It was able to spot the coronavirus in late December before Chinese authorities acknowledged the emergency.

“We are moving to surveillance efforts in the US,” said Brownstein. It is critical to determine where the virus may surface if the authorities are to allocate resources and block its spread effectively.

“We’re trying to understand what’s happening in the population at large,” he added.

Disclaimer: The author owns a small number of shares in NVIDIA and this article is intended for information purposes only.