About 2.3 million cases of breast cancer are diagnosed worldwide each year, and about 670,000 women die from the disease.
“Breast cancer is the leading cause of cancer death among women, even with screening mammography,” says Dr. Christiane Kohl of the Technical University of North Rhine-Westphalia in Aachen.
The reason is that many cases of breast cancer are not detected by mammography, at least not at an early stage. Aggressive, fast-growing tumors in particular are often not visible on mammograms, Kuhl explains. It is precisely these tumors that kill many women.
Now a new algorithm promises to redirect screening: an artificial intelligence model can assess with high accuracy, simply by analyzing data from mammograms, a person’s risk of developing breast cancer in the next five years.
In one study, women identified by the algorithm as being at high risk for breast cancer were actually significantly more likely to develop the disease than women identified by the AI as being at “normal risk.”
“These women developed breast cancer four times more often than those with low AI scores,” says Kuhl, who is the study’s lead author. “Thanks to the artificial intelligence we have developed, we can more accurately predict whether a woman will develop breast cancer in the next five years – based on a mammogram that shows no signs of the disease.”
Individual tracking
In general, regular mammograms for breast cancer screening are recommended for women aged 50 to 74 years, every two years. However, the individual risk of developing the disease – and thus the need for effective early detection – varies greatly from woman to woman.
Therefore, Kuhl advocates individual breast cancer screening. Ultimately, the accuracy of mammography varies widely from woman to woman: the denser the breast tissue, the greater the risk of developing the disease – and the worse it is recognized by mammography. The doctor says that many women do not know this.
Doctors recommend that women with very high breast density undergo magnetic resonance imaging (MRI) for early detection, a test that helps reliably identify breast cancer at an early stage. Although an MRI is very reliable, it costs several times more than a mammogram or ultrasound, which are less reliable.
Artificial intelligence can decide whether an MRI is necessary or not
To identify women who need MRI for early detection, the Clearity Consortium (an international collaboration of 46 research institutions in the United States, Canada, South America and Germany) developed the ClearityBrest AI system, which was trained on hundreds of thousands of mammograms from the Americas and Europe.
Unlike traditional risk models, the algorithm does not require information about family history, genetics or lifestyle. It calculates the probability of developing breast cancer exclusively from mammography and classifies women into risk groups based on specific thresholds.
The AI recognizes not only the amount of glandular tissue but also its texture, which is another indicator of breast cancer risk. “Only about 10% of women have this very dense glandular tissue. The vast majority of women who get breast cancer and are diagnosed late have less dense tissue,” Kuhl says.
The crucial advance, in his view, is that “AI can decide in seconds whether a woman needs an MRI for early detection or not.”
Another approach
In most countries, systematic screening for breast cancer begins at age 50, because the risk increases significantly with age and the benefits of widespread mammography are statistically proven from this age onward.
Although younger women are less likely to develop breast cancer, they are more likely to develop aggressive tumors if they develop the disease.
“In fact, younger women will especially benefit from early detection — as long as it works,” Cole says. Because mammograms tend to be problematic for younger women specifically: “Breast tissue in young women is usually dense — this makes early detection by mammography particularly difficult.”
However, Cole says that simply lowering the screening age is not very effective. Instead, it calls for a two-step approach. “First, mammograms for early detection; then AI analysis should be done to determine the risk of developing the disease in the next five years.”
If the algorithm indicates a particularly high risk, MRI should be offered, and mammography is no longer necessary for these women.