UK Study Tests for Optimal AI Workflow Implementation for Breast Cancer Detection

Monday, December 2, 2024

By Melissa Silverberg

The first large-scale prospective study on using AI in breast cancer screening in the UK shows that with the right workflow, the technology shows significant promise for improving cancer detection accuracy and service efficiency.

Lip

The results of the study presented on Sunday, mapped different workflows to find the right balance between many factors radiologists consider in AI adoption such as workload savings, reducing false positives and increasing cancer detection, said Gerald Lip, MD, clinical director, North East of Scotland Breast Screening Programme and president of the British Society of Breast Radiology. Dr. Lip also chairs an AI Special Interest Group at the British Institute of Radiology.

This evaluation aimed to compare and evaluate different workflows for integrating AI into breast screening, assessing their impact on screening outcomes and workload. The evaluation included 10,889 women between February and October 2023 who received standard double reading with arbitration of discordant cases. If the double reading process did not prompt recall but the AI flagged potential abnormalities, women were identified as needing further evaluation and were recalled.

Routine evaluation detected 106 cancers or 9.7 per 1,000. The primary workflow resulted in additional evaluation for 1,345 (12.4%) women after being flagged by the AI, yielding an increased cancer detection rate of 1 per 1000. The additional cancers detected were mostly invasive and high grade.

The team modeled 17 different workflows including two where the AI worked as an extra reader, five where AI could be used as a triage tool, and ten more with a combination of both approaches.

“This evaluation presents the advantages and challenges of different workflow options for integrating AI into breast screening. The primary workflow and multiple other variations emerge as particularly promising strategies, offering superior clinical outcomes and operational gains,” Dr. Lip said. “These findings underscore the importance of tailoring AI integration to specific clinical contexts to optimize its impact on breast screening and show that different AI implementation strategies provide improved cancer detection and workload savings with varied trade-offs.”

Staff Buy In Most Important

Dr. Lip said implementation of AI into the hospital’s workflow was not technically difficult, but that education and communication were key to getting the entire team on board from administrative staff to technicians, radiologists, surgeons and oncologists. He is confident that with proper planning, AI can be a powerful tool for radiologists in diagnosing as well as reducing workforce challenges. The primary workflow modelled showed workload savings of up to 30%.

“Ultimately for radiologists, the AI augments our practice making us more effective clinicians,” Dr. Lip said. “Along with picking up more cancers, the high negative predictive value we demonstrate in our work means that in UK and European screening programs where mammograms are read by two humans, partial substitution of one of the human readers for normal examinations can deliver real workforce benefits and result in reducing burnout.”

Access the presentation, “Optimising Breast Screening Pathway Integration: A Comparative Analysis of AI Workflows in a Prospective Evaluation,” (S5-SSBR02-6) on demand at RSNA.org/MeetingCentral

Back

© 2024 RSNA.
The RSNA 2024 Daily Bulletin is the official publication of the 110th Scientific Assembly and Annual Meeting of the Radiological Society of North America. Published online Sunday, December 1 — Friday, December 6.

The RSNA 2024 Daily Bulletin is owned and published by the Radiological Society of North America, Inc., 820 Jorie Blvd., Suite 200, Oak Brook, IL 60523.