The diagnosis of breast cancer is based on the study of images obtained from mammograms. However, in the case of dense or fibrous breasts, it is more difficult and it is often necessary to carry out other complementary tests. In fact, according to Dexeus Mujer statistics, of every 100 women who undergo mammography, 70 % require an ultrasound scan to complete the study. Of these, 2-3 will be biopsied, and, of these, only 1 case will confirm cancer. Not all images have the same level of suspicion. This means biopsies that cause a lot of stress and worry for patients.

For this reason, a team from the Gynaecological Diagnostic Imaging Service (DGI) in Dexeus Mujer has conducted a retrospective study to evaluate the effectiveness of a new system based on Artificial Intelligence in order to refine the diagnosis of breast cancer. This technology, developed by the company Koios Medical, has been authorised by the FDA and the European Medicines Agency (EMA).

To validate the efficacy of the system, a total of 403 echo-guided biopsies performed in 2019 on women of different ages were selected. These images were obtained during gynaecological check-ups performed by the DGI Service in Dexeus Mujer. According to the pathology report, 197 were classified as malignant lesions and 206 as benign. The authors chose the image that best represented the classification obtained on the BI-RADS scale and evaluated it with the Koios AI system. The results of this study evidence that the use of AI can help refine the diagnosis, as it increased the proportion of suspected malignant cases that were ultimately confirmed by biopsy in the highest risk categories: BI-RADS 4, 5 and 6.