Using A.I. to Detect Breast Cancer That Doctors Miss Hungary has become a major testing ground for A.I. software to spot cancer, as doctors debate whether the technology will replace them in medical jobs.
Doctors in Hungary are testing artificial intelligence (A.I.) software to help detect breast cancer, as the technology shows promising results in spotting signs that human radiologists may miss.
The A.I. systems are being tested on real patients at five hospitals and clinics in Hungary that perform over 35,000 screenings per year, and the country has become a major testing ground for the technology.
While some medical professionals are concerned about the potential for A.I. to replace doctors in medical jobs, early results suggest that A.I. can improve public health by detecting breast cancer at least as well as human radiologists. Clinics and hospitals in the United States, Britain, and the European Union are also testing or contributing data to help develop these systems.
A.I. technology is experiencing a surge in usage as it becomes the center of a Silicon Valley boom, with chatbots such as ChatGPT demonstrating its remarkable ability to communicate in humanlike language, although sometimes with concerning outcomes.
The breast cancer screening technology, which is based on a similar form used by chatbots and modeled on the human brain, is another example of how A.I. is infiltrating everyday life.
Despite the increasing adoption of the cancer detection technology, doctors and A.I. developers agree that numerous challenges remain. More clinical trials are necessary before the systems can be more widely used as an automated second or third reader of breast cancer screens, beyond the limited number of places currently employing the technology.
Additionally, the tool must demonstrate its ability to produce accurate results on women of all ages, ethnicities, and body types. Radiologists also stressed the need for the technology to recognize more complex forms of breast cancer and reduce false-positive results that are not cancerous.
Furthermore, the deployment of A.I. tools has sparked a debate about whether they will replace human radiologists. Technology makers face regulatory scrutiny and opposition from some doctors and health institutions. However, many experts believe that the technology will only be effective and trusted by patients if used in conjunction with trained doctors.
In the end, A.I. could be life-saving, as stated by Dr. László Tabár, a leading mammography educator in Europe, who was won over by the technology’s performance in breast cancer screening from various vendors.
He dreams of a day when women go to a breast cancer center and inquire, “Do you have A.I. or not?” Hundreds of images are analyzed each day using the technology.
Geoff Hinton, a renowned A.I. researcher, predicted in 2016 that A.I. would surpass the abilities of radiologists within five years. However, some experts, like Peter Kecskemethy, a computer scientist and co-founder of Kheiron Medical Technologies, knew that the reality of replacing radiologists would be more complex.
Kheiron Medical Technologies develops A.I. tools to aid radiologists in detecting early signs of cancer. Hinton and his team used neural networks, which mimic the human brain’s information processing, to build an image recognition system that can identify objects like flowers, dogs, and cars.
This technology is used by apps like Google Photos and personal assistants like Siri and Alexa. A.I. enthusiasts believed that the same technology could be applied to detect diseases like breast cancer in mammograms, as there were 2.3 million diagnoses and 685,000 deaths from breast cancer globally in 2020.
The New York Times’ Akos Stiller took the photo of Mr. Kecskemethy. Growing up in Hungary and spending time at one of Budapest’s largest hospitals, Mr. Kecskemethy witnessed the challenges radiologists face when searching for small malignancies in images. As they spend hours each day reviewing hundreds of images, radiologists must make life-altering decisions for their patients. Dr. Edith Karpati, Mr. Kecskemethy’s mother and a medical product director at Kheiron, acknowledges the difficulty of staying focused and finding small lesions in mammograms.
Mr. Kecskemethy and Kheiron’s co-founder, Tobias Rijken, who is an expert in machine learning, believe that A.I. should aid doctors. To train their A.I. systems, they obtained more than five million historical mammograms from clinics in Hungary and Argentina, as well as academic institutions such as Emory University. Additionally, the company, which is located in London, pays 12 radiologists to label images using special software that teaches the A.I. to recognize cancerous growth by its shape, density, location, and other factors.
The technology creates a mathematical representation of normal mammograms and mammograms with cancers from the millions of cases fed into the system. It then compares each mammogram’s abnormalities to the baseline with more granular attention than the human eye.
Last year, after a test on over 275,000 breast cancer cases, Kheiron reported that its A.I. software performed as well as human radiologists when acting as the second reader of mammography scans.
It also decreased radiologists’ workloads by at least 30%, as it reduced the number of X-rays they needed to examine. In another trial conducted at a Hungarian clinic last year, the technology increased the cancer detection rate by 13% by identifying more malignancies.
In 2021, Dr. Tabár, a radiologist whose mammogram reading techniques are widely used, tested Kheiron’s software using some of the most challenging cases of his career. In each case, the A.I. accurately identified developing cancer that had been missed by the radiologists.
“I was surprised at how good it was,” he said.
Dr. Tabár had no financial ties to Kheiron when he first tested the technology and has since received an advisory fee for providing feedback to improve the system. He also noted that he has tested systems from other A.I. companies, including Lunit Insight from South Korea and Vara from Germany, which have also shown promising results.
Kheiron’s technology was first used on patients in 2021 at MaMMa Klinika, a small clinic in Budapest. After a mammogram is completed, two radiologists review it for signs of cancer, and the A.I. then either agrees with their findings or identifies areas for further review.
At five MaMMa Klinika sites in Hungary, the A.I. has identified 22 cases of cancer that were missed by radiologists, with about 40 more under review. “It’s a huge breakthrough,” said Dr. András Vadászy, the clinic’s director, who was introduced to Kheiron through Dr. Karpati, one of the company’s founders. “If this process will save one or two lives, it will be worth it.”
According to Kheiron, their technology works best in conjunction with doctors rather than as a replacement for them.
The Scottish National Health Service will incorporate it as an additional reader of mammograms at six sites, while England’s National Health Service will deploy it in around 30 breast cancer screening sites by the end of the year.
Oulu University Hospital in Finland also plans to use the technology, and a bus will travel around Oman this year to perform breast cancer screenings using A.I. However, Kheiron’s founder, Mr. Kecskemethy, emphasized that “an A.I.-plus-doctor should replace doctor alone, but an A.I. should not replace the doctor.”
The National Cancer Institute estimates that around 20 percent of breast cancers are missed during mammogram screening.
While some doctors may be skeptical about A.I. in medicine, breast imaging specialist Dr. Constance Lehman of Massachusetts General Hospital encourages doctors to keep an open mind.
Dr. Ambrózay, a radiologist at Bács-Kiskun County Hospital outside Budapest, was initially skeptical of the technology but quickly changed her mind after using it to spot a tiny tumor in a 58-year-old woman’s mammogram that she had difficulty seeing.