Fredrik Strand

Karolinska Institute, Solna, Sweden

Fredrik Strand is a Swedish radiologist and associate professor at Karolinska Institutet, Stockholm. He holds MD, PhD and MSc(Eng) degrees. He is head of the research and education committee at the Swedish society of breast imaging. Fredrik is the principal investigator of two AI-based clinical trials: ScreenTrustCAD and ScreenTrustMRI. His research evolves around exploring AI on multimodal radiology images - for early detection, cancer characterization, treatment response and prognosis. Fredrik and his team curate high-quality datasets and radiologist annotations to develop new AI models for breast imaging models, performing reader studies and external validations. He leads the Swedish multicenter VAI-B, Validation platform for AI in Breast imaging, which is a platform hosting images from various clinical centers linked to clinical outcomes with the mission to test external AI CAD models. Fredrik is engaged in two EU-funded projects, RadioVal and EUCAIM; and he is also involved in the EUCanScreen initiative developing guidelines for safe implementation of AI in clinical practice. Fredrik's long-term aim is to explore the full power of AI on multi-modal radiology combined with human expertise to improve outcomes for women with breast cancer.

LINKS:
Personal: https://ki.se/en/people/fredrik-strand#about-me
Group: https://ki.se/en/research/research-areas-centres-and-networks/research-groups/computational-breast-imaging-fredrik-strands-group
LinkedIN: https://www.linkedin.com/in/fredrik-strand-ki/
ResearchGate: https://www.researchgate.net/profile/Fredrik-Strand
Google Scholar: https://scholar.google.com/citations?hl=en&user=q3K8X-IAAAAJ&sortby=pubdate


Implementation of AI in screening

Artificial intelligence has rapidly advanced from retrospective validation studies to prospective clinical trials, and we are now entering the phase of implementation. This talk will focus on how AI can be integrated into population-based breast cancer screening programs, using evidence from large-scale trials such as ScreenTrustCAD and MASAI, as well as real-world deployment experiences. The central question is no longer whether AI can detect cancers, but how it should be integrated into radiology workflows, how it can be trusted and monitored. Key challenges include defining optimal use cases: triage, independent reading, or double reading replacement. Important are radiologist trust, patient acceptability and robust health economic evaluation. This presentation will highlight landmark study findings, practical lessons learned in workflow integration, and the importance of calibration and monitoring. Ultimately, successful implementation requires moving beyond diagnostic performance studies to a system-level perspective where AI is continuously validated, adapted, and evaluated for real-world impact.

  


UKIBCS

Please contact the conference secretariat:
t: 01332 227776
e: ukibcs@kc-jones.co.uk