Skip to main content

Artificial intelligence gives hope to women with dense breast tissue

The traditional mammography screening cannot always find the cancer, therefore the Breast Cancer Association suggests a differentiated screening for women with dense breast tissue. Artificial intelligence will hopefully improve the chances of finding breast cancer earlier, says cancer researcher Elsebeth Lynge.

When the current mammography screening is combined with artificial intelligence - or AI as it is also called after the English artificial intelligence - there is a lot of evidence that you can both save lives and save money, says cancer researcher Elsebeth Lynge, who is professor emeritus at the University of Copenhagen and who Among other things, researches the effects of breast cancer screening.

Not a 100 percent safe solution

"The screening program we have now, where all women aged 50 to 69 are invited to a mammogram every two years, is not a 100 percent safe solution. For screened women as a whole, approx. one third of all breast cancer between screenings, and for women with very dense breast tissue, it approx. half of all breast cancers. It is therefore important to test new models for screening. I think the most promising thing at the moment is to combine the current mammography with advanced AI reading. Information can then be obtained both about signs of cancer or its precursors, and about the density and structure of the breast tissue. Obtaining all this information at the same time will be difficult and resource-intensive for the radiologists," explains the cancer researcher, who, however, believes that new methods must be tested thoroughly before deciding what to offer women with dense breast tissue.

Younger women have denser breast tissue

One of the dilemmas is that breast tissue changes with age as the mammary glands contract. So the younger women you screen, the greater the proportion of women with dense breast tissue will be. For example, 10 percent of the screened women in Sweden have dense breast tissue, because the Swedes already screen from the age of 45. Therefore, you cannot simply compare figures between countries, says Elsebeth Lynge, adding that the risk of breast cancer is much lower for younger women.

"This is one of the reasons why in Denmark we only screen from the age of 50. Furthermore, if we screen earlier, we find more false positives, i.e. findings where women are recommended further investigation, but where it turns out not to be breast cancer. It is a balancing act, because the task is to find the cancer without those who do not have cancer having to undergo extra examinations. Extra examinations are unpleasant for the individual woman, but it is also expensive for the healthcare system," she explains.

AI replaces the one radiologist

Today, two images are taken of each breast, and then two radiologists look at the images to see if there are changes that indicate cancer or pre-cancers. They look at it independently of each other and see if there is a consensus. If they disagree with the assessments, they look at it together once more.

"The decision whether there are signs of a lump or something else that could indicate a pre-stage of breast cancer is the basis for what is further recommended to the woman. That's why the radiologists go to great lengths with the assessment," explains Elsebeth Lynge.

In the Capital Region, the one radiologist is now replaced by AI, which looks for the same changes as the radiologist. If there is disagreement between the assessment by the radiologist and the AI, a consensus assessment is also made here. Before putting this version of AI into use, it was tested on historical material. That is, the AI evaluated images that had already been evaluated by radiologists. Here it was seen that AI works just as well as the radiologists in finding signs of cancer, also when it came to changes in very dense breast tissue. At the same time, the number of women with false-positive findings fell by a quarter, and almost two-thirds of the radiologists' time was saved.

Five percent have very dense breast tissue

"In the Capital Region, the radiologists have coded the images from screening for density of the breast tissue for research use. From this we know that five percent of the screened women have very dense breast tissue and that these women have a 2.5 times higher risk of getting breast cancer than women with very fatty breast tissue. Furthermore, we can see that only half of the breast cancer cases in women with very dense breast tissue are detected by screening," explains Elsebeth Lynge, who hopes that future AI technology can change that, so that a larger proportion of breast cancer cases are detected by screening.

"But we're not quite there yet," she says.

Here is the reference to the study of AI from the Capital Region;

Lauritzen AD, Rodríguez-Ruiz A, von Euler-Chelpin MC, Lynge E, Vejborg I, Nielsen M, Karssemeijer N, Lillholm M. An Artificial Intelligence-based Mammography Screening Protocol for Breast Cancer: Outcome and Radiologist Workload. Radiology. 2022 Jul;304(1):41-49. doi: 10.1148/radiol.210948. Epub 2022 Apr 19. PMID: 35438561.