arXiv 2504.202202025-04-28cs.CLcs.CV

用于临床数据提取的多模态流水线：将视觉语言模型应用于输血反应报告扫描件

A Multimodal Pipeline for Clinical Data Extraction: Applying Vision-Language Models to Scans of Transfusion Reaction Reports

Institute for Transfusion Medicine, University Hospital Essen, Hufelandstraße 55, Essen, Germany（1 输血医学研究所，埃森大学医院，Hufelandstraße 55，德国）；Department of Computer Science, University of Applied Sciences and Arts Dortmund (FHDO), Emil-Figge Str. 42, Dortmund, Germany（2 计算机科学系，多特蒙德应用科学大学（FHDO），Emil-Figge Str. 42，德国）；Institute for Medical Informatics, Biometry and Epidemiology (IMIBE), University Hospital Essen, Hufelandstraße 55, Essen, Germany（3 医学信息学、生物统计学和流行病学研究所（IMIBE），埃森大学医院，Hufelandstraße 55，德国）；Institute for AI in Medicine (IKIM), University Hospital Essen, Girardetstraße 2, Essen, Germany（4 医学人工智能研究所（IKIM），埃森大学医院，Girardetstraße 2，德国）；Institute of Interventional and Diagnostic Radiology and Neuroradiology, University Hospital Essen, Hufelandstraße 55, Essen, Germany（5 干预性和诊断放射学及神经放射学研究所，埃森大学医院，Hufelandstraße 55，德国）；Department of Dermatology, University Hospital Essen, Hufelandstraße 55, Essen, Germany（6 皮肤科系，埃森大学医院，Hufelandstraße 55，德国）

Henning Schäfer, Cynthia S. Schmidt, Johannes Wutzkowsky, Kamil Lorek, Lea Reinartz, Johannes Rückert, Christian Temme, Britta Böckmann, Peter A. Horn, Christoph M. Friedrich

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对纸质医疗文件人工转录耗时易错的问题，提出开源多模态流水线，集成复选框检测、OCR和VLMs提取分类数据，在输血反应报告上实现高精确率与召回率，减轻行政负担并促进准确监管报告。

摘要：

尽管电子健康记录的采用日益广泛，但许多流程仍依赖纸质文件，反映了医疗保健交付所处的异构现实条件。将纸质数据转换为数字格式时，人工转录过程耗时且易出错。为简化此工作流，本研究提出一个开源流水线，用于从扫描文件中提取和分类复选框数据。该方法在输血反应报告上进行了演示，其设计支持适配其他富含复选框的文件类型。所提方法集成了复选框检测、多语言光学字符识别（OCR）和多语言视觉语言模型（VLMs）。与2017至2024年逐年编制的黄金标准相比，该流水线实现了高精确率和召回率。其结果减轻了行政工作负担并实现了准确的监管报告。该流水线的开源可用性鼓励对复选框表单进行自托管解析。

英文摘要：

Despite the growing adoption of electronic health records, many processes still rely on paper documents, reflecting the heterogeneous real-world conditions in which healthcare is delivered. The manual transcription process is time-consuming and prone to errors when transferring paper-based data to digital formats. To streamline this workflow, this study presents an open-source pipeline that extracts and categorizes checkbox data from scanned documents. Demonstrated on transfusion reaction reports, the design supports adaptation to other checkbox-rich document types. The proposed method integrates checkbox detection, multilingual optical character recognition (OCR) and multilingual vision-language models (VLMs). The pipeline achieves high precision and recall compared against annually compiled gold-standards from 2017 to 2024. The result is a reduction in administrative workload and accurate regulatory reporting. The open-source availability of this pipeline encourages self-hosted parsing of checkbox forms.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。