文献速递胸部骨肌腹部DRUS

本地大语言模型生成患者友好型影像报告:质量与可读性的专家评估

Generating patient-friendly radiology reports with a local large language model: expert evaluation of quality and readability.
PubMed 文献 · Eur Radiol · 2026-10-29

该回顾性研究纳入荷兰初级保健的360份X线和超声报告,使用本地部署的Gemma-3(12B参数)生成患者友好版本,评估准确性和完整性。患者友好版结论的准确性评分高于主体报告,但临床相关幻觉发生于12.5%的报告,提示需临床监督。

要点

  • 患者友好版结论的准确性评分高于主体报告,完整性均高。
  • 临床相关幻觉发生于12.5%的报告,是主要安全顾虑。
  • 仅48.0%的联合部分同时满足质量和可读性标准。

核心结论

  • 本地部署的大语言模型可生成患者友好型影像报告翻译,避免数据云端传输,但临床相关幻觉仍是安全隐患,临床实施前需医生监督和进一步评估。

研究设计与样本

  • 回顾性研究,纳入360份荷兰初级保健的胸部、膝、腰椎、肩X线及肩、腹部超声报告,使用Gemma-3(12B参数)翻译,以5点Likert量表评估准确性和完整性,并计算可读性指标。

诊断标准与关键数值

  • 患者友好版结论的中位准确性评分为5(Q1-Q3 5-5),主体报告为4(3-5);两者完整性评分均为5(5-5)。
  • 患者友好版报告较原始报告更长(中位增加65词,48-84,p < 0.001),FDRE评分平均增加54.2 ± 18.8(p < 0.001)。
  • 71.0%的主体报告和96.0%的结论达到预设质量标准(准确性和完整性≥4),48.0%的联合部分同时满足质量和可读性标准(FDRE评分>60)。
  • 临床相关幻觉发生于45份报告(12.5%)。

临床应用与阅片要点

  • 报告结论部分可能是本地大语言模型生成患者友好版本的最有前景的初始应用场景。
  • 临床实施前需医生监督,并进一步开展以患者为中心的验证。

局限性

  • 回顾性研究设计。
  • 样本来自荷兰初级保健,可能限制普适性。
  • 临床相关幻觉发生率较高(12.5%),需关注安全性。

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

OBJECTIVES: To assess the feasibility of using a locally hosted large language model (LLM) to generate patient-friendly radiology reports while avoiding cloud transfer of report text.

MATERIALS AND METHODS: In this retrospective study, 360 chest, knee, lumbar spine, and shoulder X-rays, and shoulder and abdominal ultrasound reports from Dutch primary care were included. Main report sections (n = 359) and conclusion sections (n = 350) were translated into patient-friendly versions using Gemma-3 (12B parameters) and assessed for accuracy and completeness on 5-point Likert scales. Readability was evaluated using text length and Flesch-Douma reading ease (FDRE) scores. Hallucinations were systematically documented. Statistical analyses included Kruskal-Wallis, Wilcoxon signed-rank, and paired t-tests.

RESULTS: Patient-friendly conclusions achieved higher median accuracy scores than main reports (5 [Q1-Q3 5-5] vs 4 [3-5]), while completeness scores were high for both sections (5 [5-5]). Compared with original reports, patient-friendly reports were longer (median increase 65 words [48-84], p < 0.001) and had higher FDRE scores (mean increase 54.2 ± 18.8, p < 0.001). Overall, 71.0% of main reports and 96.0% of conclusions met predefined quality criteria (accuracy and completeness ≥ 4), while 48.0% of combined main report and conclusion sections met both quality and readability criteria (FDRE score > 60). Clinically relevant hallucinations occurred in 45 reports (12.5%).

CONCLUSION: A locally hosted LLM demonstrated feasibility for generating patient-friendly translations of Dutch X-ray and ultrasound reports while avoiding cloud transfer of data, with better performance for conclusion sections than main reports. However, clinically relevant hallucinations remain a safety concern and indicate the need for clinician oversight and further evaluation before clinical implementation.

KEY POINTS: Question Can a locally hosted LLM generate patient-friendly radiology reports while preserving data privacy and maintaining acceptable accuracy, completeness, and readability? Findings A locally hosted LLM generated patient-friendly report conclusions with higher quality scores than main reports, while clinically relevant hallucinations remained an important safety concern. Clinical relevance A locally hosted LLM has potential to generate patient-friendly radiology report translations without external cloud data transfer. Report conclusions appear to be the most promising initial application, although clinician oversight and further patient-centered validation remain essential before clinical implementation.

来源:Eur Radiol,查看原文。版权归原作者所有。

引用

  1. van der Mee FAM, Broos R, Ottenheijm RPG, et al. Generating patient-friendly radiology reports with a local large language model: expert evaluation of quality and readability. Eur Radiol. 2026. PMID: 42957801. doi:10.1007/s00330-026-12971-0

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。