Cloud Hosting

Researchers from Imperial College London and Google Health this week published newresearchthat shows DeepMind's medical AI system can outperform doctors on identifying breast cancer from X-ray images,per CNBC.

Business Insider Intelligence

The study involved training the AI algorithm on 29,000 anonymized images from breast cancer screenings of women from the UK and the US, and the results were compared against the performance of radiologists from both countries. When compared with radiologists from the US, DeepMind's AI system reduced false positive rates (when an image is falsely identified as abnormal) by nearly 6% and false negative rates (when a cancer is missed) by over 9% meaning there were more than 2,600 cases where a radiologist did not flag breast cancer, but an AI might have.

DeepMind's study joins a growing body of research that demonstrates AI's potential to transform radiology and improve patient outcomes.The FDAapprovedmore AI algorithms for radiology between 2017 and 2018 than for other use applications,perthe Regulatory Affairs Professionals Society: For example,GEearned approval for its AI-powered portable X-ray capable of slashing time to diagnosis for a collapsed lung from nine hours using traditional methods to a mere 15 minutes.

And AI has also proven to be on par or better than medical professionals in terms of accuracy: A 2018studyfrom Stanford tested an AI algorithm for radiology and found that the system was capable of performing as well or better than a team of three radiologists on 11 of 14 tested pathologies, accurately diagnosing 420 X-rays in 90 seconds while the team of live doctors took several hours to do the same.

However, research has shown that AI performs best when it complements traditional, human intelligence rather than supplants it entirely.While some, like investment mogulVinod Khosla, believe radiologists are an endangered species in healthcare, a follow-upstudyfrom many of the same researchers involved in the 2018 Stanford study suggests that "human-in-the-loop" workflows that utilize AI as a time-saving triage tool perform better than either AI or human doctors on their own.

As DeepMind works to secure widespread adoption of its AI in healthcare, it'll likely want to do more to address patient privacy concerns surrounding its parent company, Google.Google has come under fire in the last year for its use of patient health data, most notably around its secretive "Project Nightingale" tie-up with major US health system Ascension: It surfaced in November of 2019 that Google was conducting research using personally identifiable patient information unbeknownst to Ascension doctors or patients.

Given that only20%of US adults say they'd trust AI-generated health advice and that there's a growingdistrustof big tech among US consumers, DeepMind will likely need to provide more assurances to hospitals and patients regarding data sharing practices involving sensitive patient health info if it wants to attract wary provider partners.

Want to read more stories like this one? Here's how to get access:

Link:
Google's DeepMind AI outperforms doctors in identifying breast cancer from X-ray images - Business Insider UK