创建放射学教学内容:STELLA——标准化电子学习库与应用平台
研究构建并部署了供应商中立的开源网络应用STELLA,用于影像采集、组织、标注、展示与系统教学。部署31个月内,131名用户从2791名患者创建3467个教学病例,覆盖17个放射学亚专科;病例快速保存与本体标注支持互操作性,可能支持机构间内容共享及人工智能模型构建的图像标注。
要点
- STELLA通过PACS「发送」按钮传输研究实例唯一标识符,自动查询/检索DICOM图像与元数据。
- 每个病例至少需创建三个结构化元数据标签:亚专科、核心解剖、所见/诊断。
- 部署31个月创建3467个教学病例,覆盖17个亚专科,至少记录592个诊断。
核心结论
- 成功开发并部署STELLA,病例收集与教学利用快速增长;快速病例保存与基于本体的数据标注支持互操作性,可能支持机构间内容共享及人工智能模型构建的图像标注。
研究设计与样本
- 构建供应商中立的开源网络应用STELLA,以容器化方式安装于临床服务器,通过PACS发送按钮传输研究实例唯一标识符并查询/检索DICOM图像与元数据。
诊断标准与关键数值
- 每个病例要求至少创建三个结构化元数据标签(主要为RadLex术语):亚专科、核心解剖、所见/诊断。
- 可选元数据通过基于标准的可定制模板扩展,使用受控术语,包括疾病类别、详细解剖、介入操作、受训者级别和证据级别。
- 病例标题与教学要点叙述采用自由文本字段。
临床应用与阅片要点
- 病例默认在所有用户间共享。
- 可创建自定义项目和基于任务的工作清单以训练和测试学习者。
- 支持高效的图像与视频采集、幻灯片制作及批量病例上传。
- 受保护健康信息目前存储于数据库,但通过假名化隐藏。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
PURPOSE: Collections of interesting cases are at the heart of radiology education, but efficient saving and sharing of cases has always been a challenge. While numerous home-grown teaching file systems have been created, those deeply dependent on a specific PACS or computer technology are vulnerable to obsolescence and often lack rich educational metadata. To overcome these limitations, we built a powerful new vendor-neutral, open-source web application for image collecting, organizing, annotation, display, and systematic teaching called STELLA-a STandardized Electronic Learning Library and Applications.
METHODS AND MATERIALS: The software and system architecture passed hospital IT security reviews and was installed on clinical servers as a "containerized" application. A "send" button was added to PACS, which transmits the study instance unique identifier (SUID) to STELLA, launching a query/retrieve call to retrieve the DICOM images and metadata for display in STELLA. Users are required to create a minimum of three structured metadata tags (primarily RadLex terms) for each case as follows: subspecialty, core anatomy, and findings/diagnosis. A highly customizable standards-based metadata creation template extends optional metadata to numerous other classes using controlled terminology, including disease category, detailed anatomy, interventional procedures, trainee level, and level of proof. Free-text fields are employed for Case Title and a descriptive Narrative of teaching points. Cases are shared across all users by default. Custom projects and task-based worklists can be created to train and test learners. Efficient image and video capture, slide making, and batch case upload are supported. Protected health information, while currently stored in the database, is hidden using pseudonymization.
RESULTS: In the first 31 months of deployment, over 200 logins were created and 131 users created 3467 teaching cases from 2791 patients across 17 radiology subspecialty categories. 23 users contributed at least 10 teaching files, with the top five users contributing 1586, 553, 334, 280, 219 cases. The top six specialties represented include MSK (1793 cases), Chest (535), breast (377), abdominal imaging (183), neuroradiology (165), and pediatric radiology (159). At least 592 diagnoses were recorded across 67 core anatomy terms. Multiple organized projects and task-focused worklists have been created for systematic content delivery and testing.
CONCLUSION: We successfully developed and deployed STELLA and are experiencing rapid growth in case collection and utilization for teaching and systematic "mastery" learning. Rapid case saving and ontology-based data labeling also enable interoperability, which could support inter-institutional content sharing for teaching and learning, research, or for image annotation supporting artificial intelligence model building.
引用
- Beaulieu CF, Do B, Lin M, et al. Creation of Radiology Teaching Content with STELLA-A STandardized Electronic Learning Library and Application Platform. Acad Radiol. 2025. PMID: 41455626. doi:10.1016/j.acra.2025.11.040
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。