The Use of AI to Safely Reduce the Workload in Breast Cancer Screening With Mammography in Region Östergötland

Overview

The overall aim of the project is to investigate how artificial intelligence (AI) can be used to streamline and at the same time increase diagnostic safety in breast cancer screening with mammography. AI has been shown in a number of studies to have great potential for both increasing diagnostic certainty (e.g. reduced occurrence of interval cancers) and at the same time reducing the workload for doctors. However, much research remains to clinically validate these new tools and to increase the understanding of how they affect the work of doctors. The specific goal of the project is to investigate whether the implementation of AI in breast cancer screening in Östergötland, Sweden, can increase the sensitivity (the mammography examination's ability to find breast cancer) and the specificity (that is, the right case is selected for further investigation: a minimum of healthy women are recalled but so many breast cancer cases that are possible are selected for further investigation) and at the same time make screening more efficient through reduced workload. AI will be implemented in the clinical routine and performance metrics such as cancer detection rate etc will be closely monitored. The study do not assign specific interventions to the study participants.

The overall aim of the project is to study whether the use of artificial intelligence can improve breast cancer screening with mammography. AI will be implemented in the clinical routine and performance metrics such as cancer detection rate etc will be closely monitored. The study do not assign specific interventions to the study participants. The specific objective is to investigate whether the use of AI leads to increased diagnostic safety in mammography in Östergötland (measured as a reduced incidence of interval cancer) and at the same time leads to a reduced workload for the breast radiologists. Furthermore, the intention is to investigate how the use of AI affects the breast radiologists´ work in terms of reading time per examination and whether the radiologists' specificity and sensitivity are affected when they have access to the decision support based on AI during the review compared to if they do not have this support. The hypotheses are that: 1. The use of AI in breast cancer screening in Östergötland Sweden improves the diagnostic quality. As a result, more breast cancer cases are detected early and the incidence of interval cancer decreases. 2. The reduced workload for the radiologists in Östergötland that could be demonstrated through the data collected in Östergötland 2021-2022 \[NCT05048095 - Artificial Intelligence in Breast Cancer Screening in Region Östergötland Linkoping (AI-ROL)\] can also be demonstrated in a large-scale prospective study. 3. Through the use of an AI-based decision support, not only can double review be eliminated for those cases where the AI assesses the cancer risk as low, but also each examination can be reviewed more quickly while maintaining or improving diagnostic certainty. 4. It is the least experienced radiologists who are most helped by the decision support, both for increased diagnostic certainty and increased efficiency.