In a major breakthrough, tech giant Google is now putting Artificial Intelligence at work to identify breast cancer with greater accuracy.
According to reports, the new AI-based system scans through mammogram images to spot cancer growth, and till now has been proven to cut down false negatives by 9.5%.
This comes as a significant improvement upon the present accuracy of the tests which fail to detect breast cancer 20 per cent of the time.
Shravya Shetty, a researcher at Google and co-author of this the study was quoted in a report saying,”mammograms are very effective but there’s still a significant problem with false negatives and false positives”.
In this research, which is funded by Google, scientists took anonymous mammogram images of 25,000 women from the UK and 3,000 from the US.
The samples were fed to the AI, which was trained to scan X-ray images. The system scanned for signs of malignancy through all the 28,000 pictures, after which its findings were matched against the actual cancer status of the women.
The Verge reported that, in the US, false-negative diagnoses went down by 9.4 per cent and false positives by 5.7 per cent. False negatives and false positives went down by 2.7 and 1.2 per cent respectively in the UK where the mammograms are cross-checked by two radiologists.
The system didn’t function with spot-on accuracy all the time, as there were cases where the AI flunked in spotting cancer which was later pointed out by doctors.
“We’re very excited and encouraged by these results. There’s obviously quite a bit of nuance when you put this into clinical practice,” Daniel Tse, a Google product manager and one of the co-authors was quoted in a report.
Tse and his team are currently trying to ensure that the study’s results can be generalized for a larger population.
Google has portrayed this project as something that would complement the diagnostic skills of radiologists and not displace them from their profession. Shetty explained
“There are a number of cases where the radiologists catch something that the model misses, and vice versa. Bringing the two together could strengthen the overall results.