视觉大语言模型在成像诊断中的机遇和挑战:从临床MRI中大脑转移检测的教训
Christian Nelles1, Nour Abou Zeid1, Robert Terzis1
1Institute for Diagnostic and Interventional Radiology, Faculty of Medicine, University Hospital Cologne, University of Cologne, 50937 Cologne, Germany.
Diagnostics (Basel, Switzerland)
|March 14, 2026
概括
像GPT-4o和Claude Sonnet 3.5这样的视觉大语言模型 (vLLMs) 在MRI上检测大脑转移具有很高的灵敏度,但具有较低的特异性和幻觉,限制了当前的临床使用.
科学领域:
- 人工智能在医学中的应用
- 医学成像分析 医学成像分析
- 放射学 放射学是一门学科.
背景情况:
- 评估视觉大语言模型 (vLLMs) 的诊断准确性,以检测大脑转移.
- 评估GPT-4o和克劳德·索内特3.5使用组合成像和文字MRI数据.
- 调查癌症患者常规MRI扫描中的vLLM性能.
研究的目的:
- 确定GPT-4o和Claude Sonnet 3.5在识别大脑转移中的诊断准确度.
- 为了比较这些vLLM在灵敏度,特异性和其他诊断指标方面的性能.
- 评估vLLM在产生大脑转移的放射性报告中的可靠性.
主要方法:
- 对77名患者的100次MRI检查 (50例有,50例没有脑转移) 的回顾性分析.
- 在GPT-4o和Claude Sonnet 3.5.5中输入每个序列的单个代表切片与临床史.
- 评估vLLM生成的报告以检测准确性,过度诊断,序列识别,定位,横向性和尺寸估计.
主要成果:
- 两种vLLM都实现了100%的灵敏度,但特异性非常低 (GPT-4o: 8%,索内特3.5: 4%).
- 诊断准确性很低 (GPT-4o: 54%,索内特3.5: 52%),两种模型在12%的病例中都有幻觉病变.
- 序列识别很高,但解剖学定位和病变横向性显示出局限性.
结论:
- GPT-4o和Claude Sonnet 3.5对大脑转移的检测具有很高的灵敏度,但由于特异性低和幻觉,它们目前不适合临床使用.
- 在vLLM的空间可靠性和诊断行为需要进一步的研究,特别是关于视觉和文本输入的平衡.
- 未来的研究应该专注于提高vLLM的特异性和可靠性,以提供准确的放射性报告.
相关概念视频
Imaging Studies VII: Vascular Imaging
441
DefinitionRenal angiography, also known as renal arteriography, is an imaging technique used to obtain a comprehensive view of blood flow and the vascular structure of blood vessels in the kidneys and surrounding areas.PurposeRenal angiography detects blood vessel abnormalities in the kidneys, such as aneurysms, stenosis, thrombosis, vascular tumors, and renal artery stenosis. It evaluates kidney function and guides interventional treatments like angioplasty or stent placement.Pre-Procedure...
441
Imaging Studies I: CT and MRI
1.1K
Introduction: MRI and CT scans are crucial advancements in medical imaging techniques, playing a vital role in diagnosing conditions related to the gastrointestinal (GI) system. Each scan serves distinct purposes, targets specific areas, and requires unique nursing duties.
Description of the Procedures
Computed Tomography (CT) scan:
Computed Tomography (CT) scans use X-ray technology to generate detailed images of bones, organs, and tissues. During the scan, the patient lies on a moving table...
Description of the Procedures
Computed Tomography (CT) scan:
Computed Tomography (CT) scans use X-ray technology to generate detailed images of bones, organs, and tissues. During the scan, the patient lies on a moving table...
1.1K

