How AI is being used to detect cancers doctors miss
In a dark room at the Bács-Kiskun County Hospital outside Budapest, Dr. Éva Ambrózay, a radiologist with more than two decades of experience, stares at a computer monitor showing a patient’s mammogram.
Two radiologists had previously said that the X-rays did not show any signs that the patient had breast cancer. But Dr. Ambrózay is taking a closer look at several areas circled in red on the scan that artificial intelligence software has flagged as potentially cancer-causing.
“It’s something,” she said. She quickly ordered the woman to be recalled for a biopsy, which will be done within the next week.
Advances in artificial intelligence are starting to lead to breakthroughs in breast cancer screening by detecting signs doctors miss. So far, the technology has demonstrated an impressive ability to spot cancer, at least as well as human radiologists, according to early results and radiologists, the most recent study of how artificial intelligence can improve public health. One of the obvious signs.
Hungary has a strong breast cancer screening program and is one of the largest testing grounds for the technology on real patients. At five hospitals and clinics that perform more than 35,000 screenings a year, artificial intelligence systems, rolling out starting in 2021, can now help check for signs of cancer that radiologists may overlook. Clinics and hospitals in the US, UK and EU are also starting to test or provide data to help develop these systems.
Use of AI is growing as the technology becomes the epicenter of Silicon Valley prosperity, with the release of chatbots such as ChatGPT demonstrating how AI has the remarkable ability to communicate in human prose — sometimes with worrying results. Based on a similar format used by chatbots modeled on the human brain, the breast cancer screening technology demonstrates other ways artificial intelligence can infiltrate everyday life.
Doctors and AI developers say there are still many hurdles to the widespread use of cancer-detection technology. Beyond the limited number of places where the technology is now used, additional clinical trials will be needed before the system can be used more broadly as a second or third automated reader for breast cancer screening. The tool must also show that it can produce accurate results for women of all ages, races and sizes. Radiologists say the technology must prove that it can identify more complex forms of breast cancer and reduce false positives for non-cancers.
The AI tools have also sparked a debate over whether they will replace human radiologists, and makers of the technology have faced regulatory scrutiny and resistance from some doctors and health agencies. For now, those concerns appear overblown, with many experts saying the technology will only be effective and trusted by patients when used in partnership with trained physicians.
Ultimately, AI can save lives, says Dr. László Tabár, a leading mammography educator in Europe, who said he was moved by the technology after reviewing its performance in breast cancer screening .
“I dream of the day when women go to a breast cancer center and they ask, ‘Do you have artificial intelligence?'” he said.
hundreds of images per day
In 2016, Geoff Hinton, one of the world’s leading artificial intelligence researchers, believed that the technology would surpass the skills of radiologists within five years.
“I think if you’re a radiologist, you’re like Wile E. Coyote in the cartoon,” he said. tell the new yorker in 2017. There is no ground underneath. “
Mr. Hinton and two of his students at the University of Toronto built an image recognition system that can accurately identify common objects like flowers, dogs and cars. The technology at the heart of their system — called a neural network — is modeled on how the human brain processes information from different sources. It’s used to recognize people and animals in images posted to apps like Google Photos, and it allows Siri and Alexa to recognize what people say. Neural networks are also driving a new wave of chatbots, such as ChatGPT.
Many AI evangelists believe the technique could easily be applied to detect diseases such as breast cancer in mammograms. According to the World Health Organization, there were 2.3 million breast cancer diagnoses and 685,000 breast cancer deaths in 2020.
But not everyone thinks switching radiologists will be as easy as Hinton predicted. Computer scientist Peter Kecskemethy, co-founder of Kheiron Medical Technologies, a software company developing artificial intelligence tools to help radiologists detect early signs of cancer, knew the reality was more complicated.
Mr. Kecskemethy grew up in Hungary and worked in one of the largest hospitals in Budapest. His mother was a radiologist, and he saw firsthand the difficulty of spotting small malignant tumors in images. Radiologists often spend hours each day in dark rooms reviewing hundreds of images and making life-changing decisions for patients.
“It’s easy to miss tiny lesions,” said Dr. Edith Karpati, Mr. Kecskemethy’s mother, now Kheiron’s director of medical products. “It’s impossible to stay focused.”
Mr Kecskemethy, along with Kheiron co-founder Tobias Rijken, a machine learning expert, said AI should assist doctors. To train their AI system, they collected more than 5 million historical mammograms of patients with known diagnoses, provided by clinics in Hungary and Argentina, as well as academic institutions such as Emory University. The London-based company also pays 12 radiologists to label images with special software that teaches the AI to spot cancerous growths based on shape, density, location and other factors.
From the millions of cases fed into the system, the technology creates mathematical representations of normal mammograms and cancer patients. With the ability to look at each image at a finer granularity than the human eye, it then compares that baseline to find anomalies in each mammogram.
Last year, after testing more than 275,000 breast cancer cases, Kheiron Report Its AI software matched the performance of human radiologists when acting as a second reader for mammogram scans. It also reduces the workload of radiologists by at least 30%, as it reduces the number of X-rays they need to read. Among other results from Hungarian clinics last year, the technique increased cancer detection by 13 percent as more malignancies were found.
Dr. Tabár, who piloted the software in 2021, retrieved some of the most challenging cases of his career in which radiologists missed signs of cancer development. In every instance, the AI spotted it.
“I was blown away by how well it performed,” Dr. Tabár said. He said he has no financial ties to Kheiron, and other AI companies, including Lunit Insight from South Korea and Vara from Germany, have also provided encouraging test results.
Proof in Hungary
Kheiron’s technology will be used on patients for the first time in 2021 at a small Budapest clinic called the MaMMa Klinika. After the mammogram is done, two radiologists check for signs of cancer. The AI then either agrees with the doctor or flags the area for reexamination.
Across the five MaMMa Klinika sites in Hungary, 22 cases have been documented since 2021 in which the AI identified cancers missed by radiologists, and about 40 more are under review.
“This is a huge breakthrough,” said MaMMa Klinika Director Dr. András Vadász, who was introduced to Kheiron through Mr. Kecskemethy’s mother, Dr. Karpati. “If the process saves a life or two, it will be worth it.”
Kheiron said the technology works best alongside doctors, not instead of them. The NHS in Scotland will use it as an add-on reader for mammograms at six sites, and by the end of the year it will be in around 30 breast cancer screening sites operated by the NHS in England.Oulu University Hospital in Finland also plans to use the technology, and this year a bus will drive around Oman using AI for breast cancer screening
“AI plus doctors alone should replace doctors, but AI should not replace doctors,” Mr Kecskemethy said.
The National Cancer Institute has estimated About 20 percent of breast cancers are missed on screening mammograms.
Constance Lyman, professor of radiology at Harvard Medical School and chief of breast imaging and radiology at Massachusetts General Hospital, urged doctors to keep an open mind.
“We’re not irrelevant,” she said, “but some tasks are best done with computers.”
At the Bács-Kiskun County Hospital outside Budapest, Dr Ambrózay said she was initially skeptical about the technology — but was soon convinced. She took X-rays of a 58-year-old woman with a tiny tumor that the AI found, which Dr. Ambrózay had a hard time seeing.
The AI saw something, she said, “that just seemed to come out of nowhere.”