Clinical evidence

Scandinavian Perspective of AI in Breast Screening EUSOBI 2022 Summary and Video

22nd November 2022

Scandinavian Perspective of AI in Breast Screening EUSOBI 2022 Summary and Video


The Scandinavian key leaders of AI in breast imaging presented their research at the ScreenPoint symposium at EUSOBI 2022 in Malmo/ Sweden.

Kristina Lang presented the MASAI trial, the first prospective randomized controlled trial on the use of AI in breast screening as an alternative for double reading. Based on her previous retrospective studies, she is convinced that AI could lead to a more efficient and more effective screening programme. In the MASAI trial at Unilabs/Skane University Hospital Malmo, women are randomly assigned to a control arm where exams are double read as usual, or to the AI-based intervention arm: Transpara triages screening exams based on risk for malignancy and assigns 90% of all screening cases to single reading, and 10% to double reading. In addition, the top 1% most suspicious cases are automatically recalled. Almost 100k women already participated – we are excited to hear that the MASAI trial is close to completion!

“AI could lead to a more efficient and more effective screening programme.”

Solveig Hofvind presented the evidence supporting the use of AI in BreastScreen Norway. In preparation for a multi-site prospective RCT across Norway, Solveig Hofvind and her team tested several implementation strategies of AI in a retrospective simulation on over 120k screening exams. Her work provides a comprehensive overview on possible AI implementation strategies and expected outcomes regarding workload reduction, recall rates and cancer detection rates (recently published as Larsen et al., 2022 Eur Radiology). In 2023, Solveig Hofvind will start a nationwide multi-site prospective RCT, including 150k women annually. Screening exams with a Transpara score of 1 to 5 will be single read, while exams with score > 5 will be double read as usual. Based on their retrospective work, the researchers anticipate that they will reduce reading volume by 25% whilst operating at the same sensitivity for screen detected cancers. Additionally, they expect to benefit from earlier detection of interval cancers as seen in the earlier studies.

Ilse Vejborg already implemented AI in the screening programme of Denmark’s capital region with the main goal of reducing the radiologists’ workload while keeping the quality indicators stable. The promising results of their retrospective study on over 100k screening cases laid the ground for this innovation (recently published as Lauritzen et al., 2022 Radiology). In this retrospective simulation, Transpara triaged all screening exams based on their risk level. Only intermediate to high risk cases were double read, while low risk cases were dropped from human reading. An additional security net automatically recalled the most suspicious cases. The sensitivity of such an AI-based workflow would allow for 62.6% workload reduction with non-inferior sensitivity and higher specificity compared to standard double reading – Moreover, avoiding 25% of false positive recalls! These results convinced Ilse Vejborg and her team to implement AI in their clinical workflow. Since November 2021, Transpara has been prospectively implemented in the screening program of Denmark’s capital region. Each screening exam is first analyzed by Transpara. Low risk cases (score 1-7) are single read while intermediate to high risk cases are double read with Transpara as reading aid. Since the prospective implementation of Transpara, screening workload reduced by 31% and the recall rate dropped from 3% to 2.25%.

“A prospective implementation of AI in screening demonstrates a 98.81% agreement between Transpara and radiologists for low risk cases.”

“Transpara can detect cancers, the readers did not find.”