‘Second eyes’: AI-assisted breast cancer screening can detect more cancers earlier, groundbreaking trial finds

A first-of-its-kind trial demonstrates that AI-assisted mammography can improve outcomes for breast cancer patients, especially those with advanced disease.

Although many people have only recently begun using artificial intelligence (AI) in their daily lives, the use of this technology in healthcare began about a decade ago, particularly in the field of image-based diagnostics. Researchers have trained AI programs to recognize signs of tumors and other diseases from a variety of medical images, including X-rays, MRIs, and tissue biopsies on slides.

These images were primarily taken in “retrospective” studies of patients whose diagnosis was already known. Knowing the diagnosis allows doctors to provide feedback to the AI ​​to confirm that the AI ​​has correctly identified the cancer or that the image is free of cancer. These studies show that AI has the potential to become an invaluable tool in diagnostic medicine.

you may like

But “prospective” studies are needed to know if AI tools can truly diagnose cancer and make a difference for patients. The study uses AI tools to track diagnosed patients over several years to determine their health status.

Now, researchers in Sweden have conducted a gold standard trial to evaluate the use of AI in mammography screening. Results from the Mammography Screening with Artificial Intelligence (MASAI) trial, published in the journal The Lancet on January 31, showed that mammography interpretation supported by AI can improve screening performance while reducing the workload of radiologists.

This is the first time that an AI has been shown to improve outcomes for breast cancer patients.

Detect cancer early

Regular testing of patients has significantly reduced the incidence of late-stage cancer and breast cancer deaths in many parts of the world. However, some cancers may go undetected even if you have regular mammograms.

These “interval cancers” are not detected at the initial screening but are diagnosed within the next two years or between two screenings. These are often missed because they are masked during the initial screening by breast tissue density or tumors masquerading as normal tissue. Or, in some cases, the onset may occur rapidly between screening dates.

These cancers are invasive, spread to nearby healthy tissue, are usually aggressive, and worsen patient outcomes. Reducing interval cancer rates is the best way to ensure that screening methods are working. In other words, by detecting more cases early, we can reduce the number of terminal cancer diagnoses.

“If you want to improve the effectiveness of screening, interval cancer incidence is a very good surrogate measure of breast cancer mortality,” Dr. Kristina Lang, breast radiologist and clinical lead author at Lund University in Sweden, told Live Science. “Therefore, if we can lower the interval between cancers, it will likely have a positive impact on patient outcomes.”

you may like

The MASAI trial involved more than 100,000 women aged 40 to 80 living in Sweden. It used a commercially available AI system trained on more than 200,000 tests conducted by medical institutions around the world.

In the comparison group, mammograms were read by two radiologists, as is the Swedish standard. In the AI ​​support group, an AI system analyzed suspicious findings on mammograms and provided a risk score from 1 to 10. Cases with a score of 1 to 9 are then interpreted by one radiologist, and cases with a score of 10 are interpreted by two radiologists. The AI ​​system can also highlight suspicious findings in images, making them easier for human radiologists to examine.

AI-assisted screening identified more clinically relevant cancers than unassisted mammography. A “clinically relevant” cancer is one that has the potential to progress and therefore requires medical intervention.

The interval between cancer diagnoses within two years after the test also decreased. This shows that AI programs can more effectively identify cancers that human radiologists would normally miss, allowing treatment to begin earlier.

Reducing false positives

Although cancer screening is almost always beneficial, there are some potential downsides, such as false positives and overdiagnosis. If a patient is called in for a re-examination after being screened, but doesn’t have cancer, “it can be a very stressful experience,” Lang says.

The latter term, overdiagnosis, refers to situations in which screening detects cancers that ultimately do not harm the patient. These cancers grow very slowly and do not cause symptoms or increase the chance of death during the patient’s lifetime. Overdiagnosis can result in healthy patients receiving unnecessary cancer treatment.

The goal of AI-assisted mammography is to improve the ability of screening tests to detect cancer while mitigating these potential negative effects. This study found that AI-assisted screening did not increase the risk of false positives and improved detection of clinically relevant cancers.

In addition to improving cancer detection, AI-assisted screening could also address the persistent shortage of radiologists who can provide cancer screening.

“In some places, you’re lucky to have one radiologist available to read your mammogram,” says Dr. Richard Wall, a radiation oncologist at Washington University in St. Louis. “Without a specialist radiologist, women are not able to benefit from screening programs as much as they should.”

Furthermore, longer working hours for a small number of radiologists reduces their capacity. However, the AI ​​never tires, and its performance does not degrade at the end of the working day.

“The workforce issue is real; [study] “It could have an impact,” Wahl said, “because people will gradually become interested in AI-assisted interpretation as a second set of eyes.”

Lang and her team plan to begin a screening trial in Ethiopia in March, during which they will leverage AI to support rapid assessment of breast cancer using bedside ultrasound within screening programs.

“The problem with these facilities without a screening program is that a lot of women come in with terminal illnesses and they don’t have radiologists,” Lang said. With the support of AI, Lång hopes to improve access to accurate screening and enable early diagnosis of breast cancer in resource-limited settings.