WHAT’S BEING CLAIMED:
- A partnership between Google and medical doctors led to the creation of a deep-learning algorithm that can spot cancer signs in medical screenings with more precision than medical professionals.
- During the study, the computer brain successfully attained an accuracy of 94.4% after it was assigned to assess over six thousand cancer screening scans.
- Whereas, when pitted against human brains of radiologists without providing supplemental data, the algorithm’s performance was found to have surpassed those of the professionals.
While it’s possible that giving too much power to machines built by humans may end badly, AI (Artificial Intelligence) seems to be establishing itself as a potential tool to make lives better. One such AI, as reported by the New York Times, is found to be doing very remarkable things.
Based on an algorithm developed by Google and medical professionals, an incredible accuracy at detecting lung cancer has been demonstrated in the newly published research. In some ways, AIs have shown more accuracy at recognizing the telltale signs of cancer in medical scans than trained radiologists.
To test deep-learning algorithms for use in medical fields, the study used scans from past lung cancer screenings to train the computer brain. To determine how precisely it could identify cancers that are already known by doctors to be present, the computer brain was tasked to read more than 6,700 cancer screening scans. An impressive 94.4% accuracy was reached in this round of testing.
The next step involved assessing how skilled both the computer and the humans were at spotting cancer in scans they had never seen before.
Set against the human brains of six radiologists, the algorithm performed similarly to the professionals during the screening of data where tomography scans were provided for the radiologists. But in cases where supplemental tomography data was not available, the AI easily outdid its 6 human counterparts with 11% fewer false positives and 5% fewer false negatives.
Although this may sound like bad news for doctors, the goal here is doctors will always be part of the equation and not to be replaced by ultra-computers especially with something as serious as cancer screenings. But if by merging a human doctor’s experience with a deep-learning algorithm, then lesser mistakes in diagnosis will likely be made, thus, a better quality of life for everyone.
The paper was recently published in Nature Medicine.
Source: New York Post