Your browser doesn't support javascript. This means that the content or functionality of our website will be limited or unavailable. If you need more information about Vinnova, please contact us.

Our e-services for applications, projects and assessments close on Thursday 25 April at 4:30pm because of system upgrades. We expect to open them again on Friday 26 April at 8am the latest.

AI in breast cancer screening

Reference number
Coordinator Karolinska Universitetssjukhuset - Funktionsområde thoraxradiologi
Funding from Vinnova SEK 1 895 984
Project duration February 2017 - March 2019
Status Completed
Venture Digital health

Purpose and goal

To develop and test an AI-system that analyzes mammography images and which can be used to improve the breast cancer screening process. To achieve this, we have built a a database with mammography images in combination with clinical data from the breast cancer registry. The database serves to train and test algorithms of artificial intelligence. We have developed a decision support system (an algorithm) to determine the short-term breast cancer risk. We have tested this decision support system in a simulated clinical environment to show that it is doable.

Expected results and effects

Today there is no validated AI-system with enough precision to be useful in Swedish mammographic screening, but there is rapid progress in the field. After additional training and fine-tuning our decision support system can lead to economic value through a diagnostic process with higher accuracy and efficiency. Less women would die in breast cancer if their tumor is detected earlier. Also, the digital decision support system can easily be distributed to all regions and thereby reduce regional inequalities.

Planned approach and implementation

We have collected images and clinical information for all females within SLL:s screening program and breast cancer care 2008-2016. We have developed a validated database that now can be used as training material or the ground truth for testing algorithms. We have trained our own developed deep learning networks. Sectra AB has assisted in integration with an algorithm into a clinical test environment at KS to prepare for a prospective clinical verification. We have besides our own algorithms also assisted other external parties with validation of their algorithms.

The project description has been provided by the project members themselves and the text has not been looked at by our editors.

Last updated 5 April 2019

Reference number 2017-01382

Page statistics