- The new form of coronavirus named Covid-19 is spreading fast.
- inferVISION, a Beijing-based startup, has developed a GPU-accelerated AI tool to identify visual signs of Covid-19.
- It has already been used to examine over 32,000 cases at 34 hospitals in China.
What started in a small ‘wet market’ in Wuhan, China (which sold both live and dead animals, including birds and fish) has now become a major global issue.
Identified as a new form of coronavirus, Covid-19 is spreading fast. As of the first week of March, more than 92,000 cases and over 3,150 deaths have been registered.
To fight this outbreak, inferVISION has developed a GPU-accelerated AI tool to identify visual signs of Covid-19. inferVISION is a Beijing-based startup that develops medical AI solution.
A team of physicians at the Zhongnan Hosptial is using their software to spot likely cases of this highly contagious virus almost immediately. This is helping them isolate patients with higher efficiency, instead of waiting days for more traditional tests.
How This Tool Works?
To detect signs of Covid 19 infections, the tool looks for pneumonia caused by the coronavirus. inferVISION already had a pneumonia model before the outbreak started. They configured this model to precisely analyze coronavirus pneumonia.
It was trained on over 2,000 CT scans from early coronavirus patients. The core of the tool lies in NVIDIA Clara, an application framework optimized for Healthcare and Life Sciences developers. It consists of full-stack GPU-accelerated libraries and reference applications for advanced imaging and genomics.
Before the outbreak, inferVISION systems were trained on hundreds of thousands of lung scans to detect cancer. These training images were gathered from various hospitals in China.
InferRead CT Pneumonia solution for quickly detecting Coronavirus | Credit: inferVISION
This AI-powered system has three major advantages:
1. Unlike traditional techniques that take longer to process, it allows doctors to determine the most suitable treatments rapidly. Some people, for example, have been seen without common symptoms like cough or fever but show GGO (ground glass-opacity) in the lung, which later confirmed to be Covid-19.
2. It can quickly cross-compare developments and alterations in the lung over different tests for the same patients. In this hectic outbreak environment, manually performing this task could be difficult and could lead to wrong results. Several instances have already been registered, where doctors weren’t able to spot changes in shape and volume, but AI successfully showed minor, crucial changes in density.
3. It reduces the human resource (thus possibilities of getting infected) to perform PCR diagnostic. In fact, AI and CT can serve as a surrogate to PCR diagnostic in outbreaks similar to this.
So far, the new AI tool has been used to examine over 32,000 cases of Covid-19 at 34 hospitals in China. At present, it is being evaluated by clinics in the United States and Europe.