发表机构
Technical University of Munich (TUM); TUM University Hospital; Technical University of Munich, School of Medicine and Health; Klinikum rechts der Isar; Charité – Universitätsmedizin Berlin; Freie Universität Berlin; Humboldt Universität zu Berlin; Imperial College London; Munich Center for Machine Learning (MCML); University Hospital Essen (AöR); Institute for Artificial Intelligence in Medicine (IKIM); Institute of Interventional and Diagnostic Radiology and Neuroradiology; National Center for Tumor Diseases West（慕尼黑工业大学（TUM）; 慕尼黑工业大学医院; 慕尼黑工业大学医学与健康学院; 右伊萨尔医院; 柏林夏里特医学院; 柏林自由大学; 柏林洪堡大学; 伦敦帝国理工学院; 慕尼黑机器学习中心（MCML）; 埃森大学医院（AöR）; 医学人工智能研究所（IKIM）; 介入与诊断放射学及神经放射学研究所; 西部肿瘤疾病国家中心）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
RadPRISM将放射学模式作为分层轴，通过专用视觉子空间对齐临床概念，提升零样本分类与视觉定位性能，实现可透明检查的概念解耦医学图像表示。
AI 中文摘要
视觉-语言预训练可从放射学报告中学习丰富的医学图像表示，但以往模型变体通常在单一共享嵌入空间中运行，因此需事后恢复概念级结构与可解释性，这限制了模型透明度及临床实用性。本文提出RadPRISM，将临床医生定义的放射学模式作为指定分层轴：部署的大型语言模型从自由文本报告中提取每个概念的文本片段，每个临床概念在各自专用视觉子空间中对齐，将概念分层转化为直接的顶层对齐监督。该方法在胸部X光片上实例化，采用19概念模式，基于内部多年存档的203602次检查数据，相比匹配的全局对齐基线，RadPRISM将内部数据集零样本分类的宏观AUROC从0.717（95%置信区间0.710-0.723）提升至0.868（95%置信区间0.863-0.872）；在外部零样本分类中，其表现与专用的CARZero参考模型相当，但在指游戏视觉定位任务中大幅优于该模型（最高达4.3倍）。此外，放射科医生阅读研究显示，该方法具备概念分层检索能力（前3名的宏观检索正确率为0.78），能呈现报告级检索与固定标签词汇无法表达的解耦描述性发现。RadPRISM可生成具有判别性、空间保真度且原生概念分层的表示，这些表示由临床医生塑造并可被透明检查。
英文摘要
Vision-language pretraining learns rich medical image representations from radiology reports, but previous model variants commonly operate within a single shared embedding space, so concept-level structure and interpretability must be recovered post hoc, limiting model transparency and, hence, clinical utility. We introduce RadPRISM, which makes a clinician-defined radiology schema a designated stratification axis: an on-premise large language model extracts per-concept text spans from free-text reports, and each clinical concept is aligned in its own dedicated visual subspace, turning concept stratification into direct, top-level alignment supervision. Instantiated on chest radiographs with a 19-concept schema over $203{,}602$ examinations from an internal multi-year archive, RadPRISM improved internal dataset zero-shot classification from $0.717$ (95% CI, $0.710-0.723$) to $0.868$ (95% CI, $0.863-0.872$) macro AUROC over a matched global-alignment baseline, performed on par with the purpose-built CARZero reference in external zero-shot classification while substantially outperforming it (up to 4.3-fold) in pointing-game visual grounding. In addition, a radiologist reader study demonstrated concept-stratified retrieval ability ($0.78$ macro retrieval correctness rate within rank 3), surfacing disentangled descriptive findings that report-level retrieval and fixed-label vocabularies cannot express. RadPRISM yields discriminative, spatially faithful, natively concept-stratified representations shaped by and transparently inspectable by clinicians.
