Scholarlink

Federated Learning (FL) offers a promising solution to the dual challenges of data privacy and multi-institutional collaboration in medical imaging. However, despite strong benchmark performance, FL models rarely reach routine clinical deployment. We hypothesize that this “last-mile” gap stems from a misalignment between current FL evaluation-focused on technical metrics-and the priorities of value-based healthcare (VBHC). We conduct a structured gap analysis comparing current FL practices with VBHC principles and emerging regulatory frameworks. Seven critical deployment axes are identified; six show high-severity gaps, and one a medium-severity gap. Supporting literature is limited: only one axis is backed by strong evidence, three by moderate, one by weak, and two by very weak reviews. Based on these findings and insights from real-world pilots, we propose a practical roadmap to align FL development with clinical and regulatory expectations. By identifying key evidence gaps and outlining actionable next steps, this work aims to inform translational strategies and support the deployment challenges addressed by the BRIDGE Workshop.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From Data to Value: Gaps in Federated Learning Evaluation for Clinical Deployment in Medical Imaging

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Beatriz Garcia Santa Cruz,
  • Jaleh Shoshtarian Malak,
  • Hanna Cwiek-Kupczynska,
  • Venkata Satagopam

摘要

Federated Learning (FL) offers a promising solution to the dual challenges of data privacy and multi-institutional collaboration in medical imaging. However, despite strong benchmark performance, FL models rarely reach routine clinical deployment. We hypothesize that this “last-mile” gap stems from a misalignment between current FL evaluation-focused on technical metrics-and the priorities of value-based healthcare (VBHC). We conduct a structured gap analysis comparing current FL practices with VBHC principles and emerging regulatory frameworks. Seven critical deployment axes are identified; six show high-severity gaps, and one a medium-severity gap. Supporting literature is limited: only one axis is backed by strong evidence, three by moderate, one by weak, and two by very weak reviews. Based on these findings and insights from real-world pilots, we propose a practical roadmap to align FL development with clinical and regulatory expectations. By identifying key evidence gaps and outlining actionable next steps, this work aims to inform translational strategies and support the deployment challenges addressed by the BRIDGE Workshop.