Breast Cancer Detection: AI Matches Two Radiologists

A study shows that AI can diagnose breast cancer just as well as two radiologists

An interim safety analysis of the first randomized trial of artificial intelligence-assisted mammography screening has been published in the Lancet Oncology journal on Tuesday. This study has the potential to lessen the burden of radiologists since mammography screening with AI is safe, more accurate, and efficient.

In the interim safety study, which involved at least 80,000 Swedish women, 20% more tumors were found to have been found by AI-supported screening than by the conventional double-reading of mammograms by two breast radiologists.

The study effectively shows that AI-supported mammography analysis may diagnose breast cancer just as well as two breast radiologists working together, with a nearly halved burden for screen reading.

The main author, Dr. Kristina Lng of Sweden's Lund University, stated in a statement that "these encouraging interim safety results should be used to inform new trials and program-based evaluations to address the severe radiologist shortage in many countries." However, they do not prove that AI is prepared to be used in mammography screening on their own.

"We still need to understand the implications on patients' outcomes, especially whether combining radiologists' knowledge with AI can help detect interval cancers that are frequently missed by traditional screening, as well as the cost-effectiveness of the technology," she added.

It has been demonstrated that mammography-assisted breast cancer screening improves prognosis and lowers mortality by identifying breast cancer at an earlier, more curable stage.

The final trial results examine whether the use of AI screening in interpreting mammography images translates into a decrease in interval cancers (cancers detected between screenings that typically have a poorer prognosis than screen-detected cancers) in 100,000 women followed over two years — and ultimately assess whether AI use in mammography screening is justified — are not anticipated for several years, according to the study's authors.

According to European recommendations, screening mammograms should be evaluated twice by two radiologists to guarantee high sensitivity.

But it takes more than ten years to train a radiologist who can interpret mammograms, according to research in a peer-reviewed journal. "There is a shortage of breast radiologists in many countries, including a shortfall of around 41 (8%) in the UK in 2020 and about 50 in Sweden," it stated. The use of AI as an automated second reader for mammograms has been suggested as a way to lessen this strain and increase screening precision.

"The technology has shown encouraging results in retrospective studies using AI to triage examinations to either single or double reading and by providing radiologists with computer-aided detection (CAD) marks highlighting suspicious features to reduce false negative results," the statement stated.

80,033 women between the ages of 40 and 80 who had undergone mammogram screening at four locations in south-west Sweden were randomized in a 1:1 ratio to receive either standard analysis performed by two radiologists without AI (intervention arm) or AI-supported analysis, which involved the analysis of the mammograms by a commercially available AI-supported mammogram reading system before they were also read by one or two radiologists (control arm).

First, the AI system examined the mammogram, ranking the likelihood of cancer on a scale of one to ten, with ten being the highest likelihood and one being the lowest. One radiologist completed the image's further analysis if the risk score was less than 10, while two radiologists completed the analysis if the AI system projected a risk level of 10.

In all, cancer was found in six out of every 1,000 women who underwent screening using AI, as opposed to five out of 1,000 in the case of traditional double reading without AI. This means that for every 1,000 women who had a screening, one more cancer was found.