FedPref: Federated Preference Learning for Structured Radiology Report Extraction
FedPref: Federated Preference Learning for Structured Radiology Report Extraction
FedPref:用于结构化放射学报告提取的联邦偏好学习
Radiology reports describe findings and locations in free text, but downstream search and analysis require these relations in a fixed schema. Learning this extraction requires labels that are unevenly distributed across institutions: smaller hospitals have less local evidence, and pooling data may be infeasible. 放射学报告通常以自由文本形式描述检查发现和病灶位置,但下游的搜索与分析工作需要将这些关系转化为固定的模式(Schema)。学习这种提取任务需要标注数据,而这些数据在不同机构间的分布并不均匀:小型医院拥有的本地证据较少,且汇集数据往往不可行。
We introduce FedPref: frozen public language models propose alternative JSON extractions, local annotations rank them, and sites collaboratively train compact Qwen3-8B adapters while sharing only model updates. A heterogeneous teacher pool provides cross-model contrast when repeated single-model samples collapse. 我们引入了 FedPref:利用冻结的公共语言模型提出多种 JSON 提取方案,由本地标注进行排序,各站点在仅共享模型更新的前提下,协作训练紧凑的 Qwen3-8B 适配器。当单一模型反复生成相同样本导致崩溃时,异构教师池可提供跨模型的对比学习。
On development data from six simulated hospitals with unequal data volume and disease prevalence, FedPref improves client-mean F1 by 2.49 points and worst-site F1 by 9.10 points compared with training each site in isolation, with the largest gains at the sites holding the least data. Central training on the pooled preference-pair union is 2.66 points higher on client-mean F1. 在来自六家数据量和疾病流行率不等的模拟医院的开发数据上,与各站点独立训练相比,FedPref 将客户端平均 F1 分数提高了 2.49 个点,最差站点的 F1 分数提高了 9.10 个点,其中数据量最少的站点获益最大。在汇集的偏好对并集上进行集中训练,其客户端平均 F1 分数高出 2.66 个点。
On a locked, 400-report manually validated gold test set, FedPref reaches 68.68 F1 and pooled training 71.67, preserving that same ordering. FedPref thus lets institutions with unequal, unpooled data benefit from collaboration without ever sharing reports or annotations. 在一个包含 400 份经人工验证的锁定黄金测试集上,FedPref 达到了 68.68 的 F1 分数,而集中训练为 71.67,保持了相同的排序结果。因此,FedPref 使得拥有不均衡、未汇集数据的机构能够在无需共享报告或标注的情况下,从协作中获益。