发表机构
National University of Sciences and Technology; Consultant Radiologist; Bahria University Health Sciences Campus; Faculty of Science, Ontario Tech University（国立科技大学; 顾问放射医师; 巴里亚大学健康科学校区; 安大略理工大学理学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对脑肿瘤MRI报告模型的诊断错误问题，提出NeuroFusion辅助报告模型，利用冻结分割特征恢复诊断性能，延迟更低且临床评估更优。
AI 中文摘要
一个功能完备的脑MRI报告生成器仍可能在诊断层面“沉默”。当基于医学Mistral-7B主干构建的多链思维链（CoT）报告模型在保留的队列上评估时，它将大部分脑膜瘤和几乎所有转移瘤命名为“胶质瘤”（诊断召回率为0.44/0.07）。然而，该模型并非完全未获取到正确答案：对其冻结的分割特征应用有监督线性探针，可在0.82的宏F₁值（5折交叉验证；随机水平约为0.33）下恢复三个肿瘤队列。我们提出NeuroFusion，一种辅助报告模型，它揭示这种潜在信号而非覆盖它：基于每个病灶特征的判别场分类器头，在其确定的输出上对快速的单遍“草稿-审查”解码器进行条件设置。构建在相同的Mistral主干上，该模型恢复了诊断性能（脑膜瘤0.92，转移瘤0.75），并在三个保留队列的9项文本内容比较中赢得8项（RaTEScore、RadGraph-F₁、GREEN；经Holm校正的配对BCa），第9项无显著损失，延迟降低5-6倍（每例约80秒 vs. 457秒）。受控阴性结果明确了该机制：学习到的诊断锚点若覆盖解码器而非仅提供信息，会将分布外转移瘤召回率降至0.03。语法约束解码使92.3%的记录保持模式合规，确保每个句子可进行蕴含检查（仅7.5%矛盾，而直接基线为36.8%）。在一项盲法9例试点中，两名经委员会认证的神经科医生独立对NeuroFusion在所有肿瘤类型上的评分最高，是唯一无严重错误的系统，并在9例中的8例给出最高评分的确认（6例明确，2例并列）。
英文摘要
A capable brain-MRI report generator can still be, in effect, diagnostically silent. When a multi-chain chain-of-thought (CoT) reporter built on a medical Mistral-7B backbone is evaluated on held-out cohorts, it names most meningiomas and almost all metastases "glioma" (diagnosis recall 0.44/0.07). Yet the answer is not absent from the model: a supervised linear probe applied to its frozen segmentation features recovers the three tumour cohorts at 0.82 macro-F$_1$ (5-fold cross-validation; chance $\approx$0.33). We introduce NeuroFusion, an assistive reporter that surfaces this latent signal rather than overriding it: discriminative field-classifier heads over per-lesion features condition a fast, single-pass draft-then-review decoder on their committed outputs. Built on the identical Mistral backbone, this restores the diagnosis (meningioma 0.92, metastasis 0.75) and wins 8 of 9 prose-content comparisons across three held-out cohorts (RaTEScore, RadGraph-F$_1$, GREEN; Holm-corrected paired BCa), with no significant loss on the ninth, at 5-6x lower latency ($\approx$80 vs. 457 s/case). A controlled negative result sharpens the mechanism: a learned diagnosis pin that overrides the decoder instead of merely informing it collapses out-of-distribution metastasis recall to 0.03. Grammar-constrained decoding keeps 92.3% of records schema-valid, making every sentence entailment-checkable (7.5% contradicted vs. 36.8% for the direct baseline). In a blinded nine-case pilot, two board-certified neurologists independently rated NeuroFusion highest in every tumour type, the only system with zero critical errors, and gave it the top-rated sign-off in eight of nine cases (six outright, two ties).
Comments10 pages, 2 figures, 3 tables. Accepted at MLCN 2026, a workshop held in conjunction with MICCAI 2026; to appear in Springer LNCS
