Hallucination in Medical Imaging AI: A Cross-Modality Analytical Framework for Taxonomy, Detection, and Mitigation under Regulatory Constraints
作者: Omar Alshahrani, Muzammil Behzad
分类: cs.AI
发布日期: 2026-06-11
💡 一句话要点
提出跨模态分析框架以解决医学影像AI中的幻觉问题
🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 医学影像 人工智能 幻觉现象 跨模态分析 FDA监管 模型微调 临床决策
📋 核心要点
- 当前医学影像AI系统的幻觉现象严重影响临床决策，现有的分类和检测方法尚不完善。
- 论文提出了一个跨模态的分析框架，旨在统一幻觉的分类、检测和缓解策略，并与FDA监管相结合。
- 研究表明，通用基础模型在幻觉特定基准上表现优于医学专用模型，强调了放射科医生的监督重要性。
📝 摘要（中文）
随着AI系统在医学影像领域的快速应用，其失败模式尚未完全理解，尤其是幻觉现象：即看似合理但实际上错误的输出，包括虚构的解剖结构、遗漏的发现、错误的侧别和虚构的测量。这篇论文综合了同行评审的研究、基准数据集和FDA的监管指导，提出了一个跨模态的幻觉分类、检测和缓解分析框架。研究重点在于如何统一现有的分类法、医学专用基础模型如何减少幻觉现象，以及哪些缓解策略在FDA监管下有效。研究发现，通用基础模型在幻觉特定基准上表现优于医学专用模型，表明狭域微调可能导致过拟合。同时，放射科医生的监督仍然至关重要。
🔬 方法详解
问题定义：论文要解决的问题是医学影像AI中的幻觉现象，这种现象会导致错误的临床决策，现有方法在分类和检测上存在不足，无法有效应对不同模态的幻觉问题。
核心思路：论文的核心思路是建立一个跨模态的分析框架，通过整合不同模态的幻觉分类、检测和缓解策略，提供一个系统化的解决方案，以满足FDA的监管要求。
技术框架：整体架构包括三个主要模块：幻觉分类、检测机制和缓解策略。幻觉分类模块整合了不同模态的幻觉特征，检测机制利用医学专用模型与通用模型的比较，缓解策略则结合了物理约束、思维链提示和人机协作。
关键创新：最重要的技术创新点在于提出了一个跨模态的幻觉分类框架，并发现通用基础模型在幻觉特定基准上表现优于医学专用模型，揭示了过拟合的潜在风险。
关键设计：关键设计包括对模型进行物理约束以减少幻觉现象，采用思维链提示来引导模型推理过程，并强调放射科医生在AI生成结果中的监督作用。
🖼️ 关键图片
📊 实验亮点
实验结果显示，通用基础模型在幻觉特定基准上超越医学专用模型，具体性能数据尚未披露，但研究强调了狭域微调可能导致的过拟合风险。此外，AI生成的警报中有很高比例需要专家修正，突显了人机协作的重要性。
🎯 应用场景
该研究的潜在应用领域包括医学影像分析、临床决策支持系统和AI辅助诊断工具。通过有效管理幻觉现象，可以提高AI系统在临床中的可靠性和安全性，进而推动医学影像AI的广泛应用。
📄 摘要（原文）
AI systems are being deployed across medical imaging faster than their failure modes are understood. At this point in time, the failure of greatest clinical concern is hallucination: clinically plausible but factually incorrect outputs, including fabricated anatomical structures, missed findings, incorrect laterality, and invented measurements in generated reports, with direct consequences, for example, for biopsy decisions, staging, and treatment planning. This structured narrative synthesizes peer-reviewed studies, benchmark datasets, and FDA regulatory guidance across five imaging modalities to produce a cross-modality analysis of hallucination taxonomy, etiology, detection, and mitigation. Specifically, we address three questions in this study: (1) how can existing taxonomies be unified across modalities?, (2) how do medical-specialized foundation models hallucinate less than general-purpose ones?, and (3) which mitigation strategies are effective and compatible with FDA lifecycle oversight? We note that three taxonomic frameworks together cover the imaging pipeline in a way no single framework does alone. We also highlight that general-purpose foundation models outperform medical-specialized models on hallucination-specific benchmarks, indicating that narrow domain fine-tuning can introduce overfitting-induced confabulation. At the same time, the oversight of radiologists remains essential; for instance, a very high percentage of of AI-generated flags required expert correction before clinical use. Physics-informed architectural constraints, Chain-of-Thought prompting, and human-in-the-loop safeguards each address different failure modes and is effective when combined. All findings are mapped to the FDA's Total Product Lifecycle and Predetermined Change Control Plan frameworks, which treat hallucination management as a lifecycle obligation rather than a pre-deployment checklist.