[ad_1]
When asked if artificial intelligence would bankrupt radiologists, Dr. Topol replied, "That's it, no!
The idea is to help the doctors, not replace them.
"It will make their life easier," he said. "Overall, there is a 30% rate of false negatives, which is missing. It should not be difficult to reduce this number. "
There are however potential dangers. A radiologist who misinterprets a CT scan may be harmful to a patient, but an imperfect A.I. The widely used system could hurt many, warned Dr. Topol. Before they are released to the public, systems need to be rigorously studied, with results published in peer-reviewed journals, and tested in the real world to ensure that they work as well as they do. in the laboratory.
And even if they pass these tests, they still need to be monitored for hacking or software problems, he said.
Shravya Shetty, a software engineer at Google and author of the study, said, "How do you present the results in a way that creates a climate of trust with radiologists?"
Another problem is: If an AI The system is approved by the DFA, then, as expected, continues to change with the experience and processing of more data, will its manufacturer have to ask again for its approval ? If so, how often?
The neural network of lung screening is not yet ready for the clinic.
"We are collaborating with institutions around the world to find out how technology can be productively implemented in clinical practice," said Dr. Tse. "We do not want to get ahead of ourselves."
[ad_2]
Source link