Connect with us


AI catches breast cancer more often than doctors alone

Avatar photo



A new study shows that artificial intelligence massively improves radiologists’ ability to identify and diagnose breast cancer – but is AI ready for life-and-death responsibilities?

Doctors are able to catch breast cancer more often and faster than before due to the help of artificial intelligence, a new study shows. 

Radiologists who are accompanied by AI are now able to screen for breast cancer faster and more effectively than they would with human-only power, thanks to new AI solutions which are taking the pressure off. 

A major shortage in specialists, it is said, can be eased with the introduction of reliable tech which can help fill in the gaps, and speed up the process of scanning mammograms. 

The study, which was led by German startup Vara and  published recently (July 2022) in the Lancet Digital Health, is the first to directly compare the ability of a human with the performance of AI. 

With help from radiologists at the Essen University Hospital in Germany and the Memorial Sloan Kettering Cancer Center in New York, the Vara team tested two approaches. 

One had the AI working alone to analyse mammograms, while in the other, the AI automatically distinguishes between scans it thinks look normal and those that raise a concern. 

A (human) radiologist who would then review them before seeing the AI’s assessment, and the AI would issue a warning if it detected cancer when the doctor did not.

Charles Lanlotz said: “In the proposed AI-driven process nearly three-quarters of the screening studies didn’t need to be reviewed by a radiologist, while improving accuracy overall.”

AI isn’t new in cancer detection 

But it’s not the first we’ve heard of AI winning the race with radiologists. 

The BBC reported last year how AI is now able to detect breast cancer from mammograms more accurately than a real radiologist

A wide, international team from Imperial College London and Google Health managed to design and coach a computer model on over 20,000 X-ray images, concluding that robotics were able to outperform some human doctors. 

Is AI ready for the responsibility?

The overall accuracy of AI is probably the most debated topic since it started taking on the ‘serious’ jobs such as healthcare. 

While advanced tech like this has proven to be highly efficient and accurate, we haven’t yet been able to call it flawless. 

Researchers at the University of Washington found that AI deployed to help detect disease could actually lead to diagnostic errors, as the technology has a tendency to hunt down shortcuts and may skip important factors of the diagnosing process. 

AI will assist, not replace

But it’s not as bad as it may sound. AI is not about to step up and put human health professionals out of a job – it’s simply here to assist health professionals and speed up the process of things like radiology scans, subsequently filling the gaps where there are staff shortages and long waiting lists.

Tim Simpson, General Manager of Hologic UK & Ireland  said: “Over recent years we have seen exciting developments in the use of artificial intelligence (AI) guided imaging for diagnostics and the positive impacts this could have for the NHS and patients. In particular, there is a huge opportunity to utilise the power of AI for cancer screening, to ensure earlier detection and speedy and accurate diagnosis.

“The Industrial Strategy AI Mission has set out a goal to use data, AI and innovation to transform the prevention, early diagnosis and treatment of chronic diseases by 2030. It predicts that within 15 years better use of AI and data could result in over 50,000 more people each year having their cancers diagnosed at an early rather than late stage.

He added: “While progress has been made already, there are still challenges to overcome to make sure health services have the data capabilities to make the best use of AI.”

Ron Fridman, co-founder of Mawi, creators of AI-powered wearable medical devices for cardiac disease diagnostic and monitoring, commented: 

“AI-powered solutions have tremendous potential in most healthcare domains, from image recognition and bio-signals analysis to smart assistants, as it pulls the opportunity for human error and allows you to process huge amounts of data to enable smart data-driven decisions in MedTech – that is something that a human could never achieve. But AI is still not adopted on a scale.  

“The biggest obstacle for all companies trying to incorporate AI in healthcare is regulatory compliance and the absence of raw data collection and storage processes required to validate the technology. A large amount of professionally labelled data is crucial for ML and deep learning algorithms.”

Health Tech World previously reported how a new AI tool will help doctors to predict the cancer risk in lung nodules seen on computerised tomography (CT). Read the full story. 


Trending stories