基于大语言模型从前列腺MRI报告文本推断PI-RADS v2.1:病灶层面与分区研究
本研究纳入1118例患者共1205个病灶条目,去除报告中的PI-RADS明确要素后,由ChatGPT-4o仅依据文本推断PI-RADS类别,与放射科医生参考分类总体完全一致率72.9%(κ=0.538),外周带优于移行带,PI-RADS 3准确性最低,结论为文本-only大语言模型可作质控辅助工具而非自主决策。
要点
- 文本-only大语言模型推断PI-RADS与医生分类仅中度一致(κ=0.538),不宜替代人工判读。
- 移行带一致性极低(κ=0.077),PI-RADS 3精确率与召回率最差,需谨慎对待。
- 临床高影响的降级(PI-RADS 4-5误判为1-2)罕见,仅占1.6%。
核心结论
- 文本-only大语言模型可从真实前列腺MRI报告推断医生标注的PI-RADS v2.1类别,一致性为中度,但性能依赖分区且在PI-RADS 3附近受限,尤其移行带。
- 此类模型最适合作为有监督的质控工具,而非自主决策工具。
研究设计与样本
- 单中心回顾性研究,纳入1118例患者共1205个病灶层面条目,来源于去除全部明确PI-RADS要素后的半结构化前列腺MRI报告。
- 使用ChatGPT-4o仅基于报告文本分配数值PI-RADS类别,以放射科医生标注类别为参考标准,评估完全一致率、Cohen κ及分类别指标。
- 分析按分区(外周带 vs 移行带)及折叠风险分层(1-2/3/4-5)进行,并评估人类观察者间一致性、模型内可重复性、时间稳定性及ChatGPT-4o与GPT-5.2的配对版本敏感性分析。
关键影像表现/征象
- 总体完全一致率72.9%(κ=0.538;macro-F1=61.2%),存在系统性高估倾向。
- 外周带一致性高于移行带(参考PI-RADS 3-5时κ=0.476 vs 0.077)。
- PI-RADS 3精确率与召回率最低,频繁出现双向误分类。
- 错误机制以扩散加权成像子评分错误最常见,且存在分区差异。
诊断标准与关键数值
- 总体完全一致率72.9%(κ=0.538;macro-F1=61.2%)。
- 外周带与移行带κ分别为0.476与0.077(参考PI-RADS 3-5)。
- 折叠风险分层后一致性提高(κ=0.610)。
- 临床高影响降级(PI-RADS 4-5误判为1-2)发生率为1.6%。
- 人类观察者间一致性极好(κ=0.916-0.967)。
临床应用与阅片要点
- 可将文本-only大语言模型用于报告质控初筛,但需人工复核,尤其关注移行带与PI-RADS 3病灶。
- 警惕模型系统性高估倾向及扩散加权成像子评分相关错误,报告时注意核对DWI描述。
- 临床高影响降级虽罕见,但对PI-RADS 4-5病灶仍应坚持人工判读确认。
局限性
- 单中心回顾性设计,结果外推需谨慎。
- 仅评估文本-only推断,未纳入图像信息。
- GPT-5.2在配对分析中优于ChatGPT-4o,但少数病例产生无效输出。
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: To evaluate the feasibility and limitations of real-world, text-only inference of PI-RADS v2.1 categories from prostate MRI reports using large language models, with lesion-level and zone-aware analysis.
MATERIALS AND METHODS: This single-center retrospective study included 1,205 lesion-level entries from 1,118 patients derived from semi-structured prostate MRI reports after removal of all explicit PI-RADS elements. ChatGPT-4o was prompted to assign numeric PI-RADS categories based solely on report text. Agreement with radiologist-assigned reference categories was assessed using exact agreement, Cohen's κ, and class-wise metrics. Analyses were performed overall, by zone (peripheral vs transition), and using collapsed risk strata (1-2/3/4-5). Discordant cases were reviewed to identify error mechanisms and severity. Human interobserver agreement, intra-model reproducibility, temporal stability, and a paired model-version sensitivity analysis comparing ChatGPT-4o with GPT-5.2 were also evaluated.
RESULTS: Overall exact agreement was 72.9% (κ = 0.538; macro-F1 = 61.2%), with a systematic tendency toward overcalling. Agreement was higher in the peripheral zone than in the transition zone (κ = 0.476 vs 0.077, reference PI-RADS 3-5). PI-RADS 3 showed the lowest precision and recall, with frequent bidirectional misclassification. Collapsing categories improved agreement (κ = 0.610). Incorrect diffusion-weighted imaging subscores were the most common error mechanism, with zone-specific differences. Clinically high-impact downgrades of PI-RADS 4-5 to 1-2 were rare (1.6%). Human interobserver agreement was excellent (κ = 0.916-0.967). GPT-5.2 outperformed ChatGPT-4o in paired analyses but produced invalid outputs in a minority of cases.
CONCLUSION: Text-only large language models can infer radiologist-assigned PI-RADS v2.1 categories from real-world prostate MRI reports with moderate agreement, but performance is zone dependent and limited around PI-RADS 3, particularly in the transition zone. These models are best suited as supervised tools for quality control rather than autonomous decision-making.
引用
- Kadirli K, Cantürk A. Real-world text-only inference of PI-RADS v2.1 from prostate MRI reports using large language models: a lesion-level, zone-aware study. Eur J Radiol. 2026. PMID: 41946016. doi:10.1016/j.ejrad.2026.112838
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。