Self-report. Conventional mammography has reduced mortality by up to 40 percent among participating women. However, approximately 30 percent of cancers are detected due to symptoms occurring between two screening occasions, called interval cancer. For these women, magnetic resonance imaging (MRI) may be more appropriate, as it has higher sensitivity and has been shown to increase the number of cancers detected. Currently, there is a shortage of skilled MRI personnel, and the equipment is expensive. Therefore, it is essential to find a way to target MRI screenings to a smaller proportion of women who would benefit most from them.

At Karolinska University Hospital, we conducted a randomized clinical study, Screen Trust MRI. [1]where the newly developed AI tool (AI Smart Density) scored every mammogram. The tool was developed in collaboration between research groups at Karolinska Institutet and the Royal Institute of Technology. Of the women who had a negative mammogram and received a high AI score (around 7 percent), half were randomly offered to participate in a follow-up examination using MRI. Of the 559 women who underwent MRI, a total of 36 cancers were detected, i.e. 64 per 1,000 MRIs. This is comparable to a previous clinical study [2] Breast density was used as the selection criterion, with 16.5 cancers detected per 1,000 MRIs. The sample size was roughly the same in the current study as in the previous study, about 7-10 percent of the population. Thus, AI-smart density became about 4 times more efficient. The majority of cancers detected in our study were invasive and many were multifocal, suggesting that AI is an effective tool in detecting clinically significant cancers.

The effectiveness of targeted MRI screening needs further study, but it can be assumed to be at the same level of detection as mammography, based on approximately 10 times more cancer cases detected per 1000 screenings compared to mammography (about 5 per 1000). We hope that this method will be of great importance for breast cancer screening in the future, especially for identifying high-risk individuals who might benefit from additional imaging. The method is currently at research level and will need approval from the Swedish Medicines Agency before it can be widely used.