Dense Breast: A definition that indicates that the glandular component within the breast is elevated. And it’s a double problem, because it makes mammograms harder to read and because it increases the risk of developing cancer by up to six times. This condition is common: it affects about half of women aged 40 and over. And since it affects the likelihood of getting sick – and also the type of exam recommended by European guidelines – classifying it in the right way is key.
Dense breasts increase the risk of tumors
by TIZIANA MORICONI
For this reason, increasingly precise and advanced tools are being put in place to classify breast density consistently and accurately during mammography. Among these is TRACE4BDensit, an artificial intelligence (AI) program capable of learning to automatically perform tasks typically entrusted to the visual perception of expert people. The software, featured on Radiology: Artificial Intelligencewas developed by the Italian DeepTrace Technologies, a university spin-off of the IUSS School of Pavia, in collaboration with a group of researchers from the University of Milan, the IRCCS Policlinico San Donato and the Italian Diagnostic Center, led by Francesco Sardanelli, head of of the Diagnostic Imaging Service at the IRCCS and full professor at the Milanese university.
Breast cancer, so artificial intelligence will improve diagnoses
by Tiziana Moriconi
The categories of density
In current clinical practice, breast density is mostly assessed visually, with reference to the American College of Radiology classification A, B, C, D: from class A for almost entirely adipose breasts to class D for extremely dense breasts. However, the visual evaluation is obviously subject to the variability of the observers. The researchers’ goal was therefore to create a tool that would overcome the problem of subjectivity.
Breast cancer: how to find out in advance
by Tiziana Moriconi
The eye of AI
“TRACE4BDensity can help us provide advice to women with dense breasts about the possibility of having, after a negative mammogram, additional screening with ultrasound, magnetic resonance or contrast mammography”, explains Sardanelli: “Unlike the United States, in fact , many countries around the world, including Italy, do not have legislation requiring doctors to provide women with dense breasts with a range of information about possible additional tests. This software, therefore, proves useful both for radiologists and for patients ”.
Breast pain, only in rare cases is there a tumor
by Tiziana Moriconi
The system is based on a type of sophisticated artificial intelligence (called deep learning supervised with convolutional neural networks) and was trained under the supervision of seven expert radiologists who, independently, visually evaluated 760 mammograms. In tests, it demonstrated 89% accuracy in distinguishing between dense and non-dense breasts.