How Nvidia Is Creating Domain-Specific AI To More Accurately Test For Covid

One of the keys for more effective AI applications in the medical field is making it more specific to the domain in which it is applied, which enhances both the speed and accuracy of the information outcome desired. JL

Bret Haensel reports in Fortune:

Nvidia is working to make its A.I. and computing platforms “domain specific.” Transitioning away from A.I. technology that is broadly applicable but must be adapted by customers for industry-specific use, Nvidia is focused on making products better suited to assist specific fields, such as radiology, digital pathology, genomic sequencing, and drug discovery, among others. An A.I.-based alternative to traditional versions of COVID-19 testing uses medical imaging to do lung segmentation and a CT scan, and then it does a classification of whether or not you have COVID.

Kimberly Powell, vice president and general manager of health care at Nvidia, believes her company, in effect, is building time machines.

Using artificial intelligence and supercomputing to drastically shorten the time doctors and researchers spend diagnosing diseases and developing antiviral drug treatments, Nvidia, which got its start as a chipmaker for video games, has played a key role in accelerating efforts to stop the spread of COVID-19.

“We’ve really essentially built a time machine for these researchers,” Powell says. “Instead of waiting three months to screen 12 billion potential drug candidates, we did it in 12 hours on the world’s largest supercomputer.”

In particular, the company is working to make its A.I. and computing platforms “domain specific.” Moreover, transitioning away from developing A.I. technology that is broadly applicable but must be adapted by customers for industry-specific use, Nvidia is focused on making its products better suited to assist specific fields, such as radiology, digital pathology, genomic sequencing, and drug discovery, among others.

For example, Nvidia––in partnership with the National Institutes of Health––has developed an A.I.-based alternative to traditional versions of COVID-19 testing.

“Our testing strategy here in the United States hasn’t been extremely straightforward,” Powell says. “But if you go back to technology like medical imaging, it’s a very common practice, it’s readily available, and the results are potentially instantaneous. And so we thought, Could we use the power of medical imaging to really give a second option for testing? And we did just that…We created an A.I. model with the NIH that does lung segmentation and a CT scan, and then it does a classification of whether or not you have COVID.”

Zig Serafin, president of the software company Qualtrics, also joins the podcast to discuss how his company is assisting governments with the logistical nightmares of both distributing a potential coronavirus vaccine and convincing a critical mass of people to take it.

In addition to helping hospitals and governments with tasks like tracking vaccine candidates, scheduling patient appointments, and contact tracing, Qualtrics is working to help gauge sentiment around the vaccine so it can provide officials with information about which people are averse to receiving it, why that is the case, and what can be done to encourage them to get it.

“Really, it comes down to actually understanding people––who they are; what their needs might be given what demographic they might be from; what age they might be; what their living conditions might be,” Serafin says. “And, then, how do you accommodate people in the best possible way? So, it isn’t just about the process. Sentiment is a big part of the way in which you end up taking action, the way you end up building trust and confidence.”

Nvidia and Qualtrics are both using tech to provide support throughout the pandemic in a variety of other ways.