文献速递技术与安全CTMR

GPT-4o 用于自动化语音转文字 CT 与 MRI 报告转录的多语言可行性

Multilingual feasibility of GPT-4o for automated Voice-to-Text CT and MRI report transcription.
PubMed 文献 · Eur J Radiol · 2024-11-17

这项概念验证研究让三名经验不同的读者通过 ChatGPT iOS 应用用 GPT-4o 各口述 100 份英文和德文合成放射学报告(含多部位 CT 和 MRI),评估错误类型、严重程度和纠正时间,并用 BERTScore 和 ROUGE 评估语义相似度与 n-gram 重叠。两种语言纠正时间无显著差异,但读者经验间存在差异;错误率相近,多数错误为轻微错误,且以技术性或排版错误为主;德文 BERTScore 显著更高,ROUGE 无显著语言差异。结论认为 GPT-4o 在多语言放射学报告转录方面具有潜力,未来研究应将其与现有放射学口述工具进行比较。

要点

  • 多数转录错误为轻微错误(德文 92.68%,n=114/123;英文 94.74%,n=90/95)。
  • 错误以技术性(德文 27.04%,n=43/159;英文 35.65%,n=41/115)或排版性(德文 23.9%,n=38/159;英文 27.83%,n=32/115)为主。
  • 德文 BERTScore 显著高于英文,ROUGE 指标两种语言间无显著差异。

核心结论

  • 研究显示 GPT-4o 在多语言放射学报告转录方面具有潜力,能处理英文和德文且错误极少、语义理解高。
  • 未来研究应将 GPT-4o 与现有放射学口述工具进行比较,评估性能、成本效益及在不同说话人群中的多语言能力。

研究设计与样本

  • 概念验证研究,三名经验不同的读者各口述 100 份英文和 100 份德文合成放射学报告。
  • 通过 ChatGPT iOS 移动应用使用 GPT-4o 进行语音转文字转录,报告涵盖不同解剖区域的 CT 和 MRI 扫描。
  • 评估指标包括错误类型、严重程度和纠正时间,并计算 BERTScore 和 ROUGE 评估语义相似度与 n-gram 重叠。

诊断标准与关键数值

  • 纠正时间在两种语言间无显著差异,但不同经验读者间存在差异。
  • 两种语言错误率相似,多数为轻微错误(德文 92.68%,n=114/123;英文 94.74%,n=90/95)。
  • 错误类型以技术性(德文 27.04%,n=43/159;英文 35.65%,n=41/115)或排版性(德文 23.9%,n=38/159;英文 27.83%,n=32/115)为主。
  • BERTScore 指标德文显著高于英文,ROUGE 指标两种语言间无显著差异。

临床应用与阅片要点

  • 原文未据此给出推荐。

局限性

  • 原文未报告局限性。

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

PURPOSE: Large language models (LLMs) promise to streamline radiology reporting. With the release of OpenAI's GPT-4o (Generative Pre-trained Transformers-4 omni), which processes not only text but also speech, multimodal LLMs might now also be used as medical speech recognition software for radiology reporting in multiple languages. This proof-of-concept study investigates the feasibility of using GPT-4o for automated voice-to-text transcription of radiology reports in English and German.

METHODS: Three readers with varying levels of experience each dictated 100 synthetic radiology reports in both languages using GPT-4o via the ChatGPT iOS mobile application. Reports included CT and MRI scans of various anatomical regions. Evaluation metrics included error type, severity, and correction time. BERTScore and ROUGE metrics were calculated to assess semantic similarity and n-gram overlap between dictated and original reports.

RESULTS: No significant differences in correction time between languages were found, but differences were observed between readers based on experience. Error rates were similar for both languages, with most errors being minor (92.68 %, n = 114/123 German; 94.74 %, n = 90/95 English) and technical (27.04 %, n = 43/159 German; 35.65 %, n = 41/115 English) or typographical (23.9 %, n = 38/159 German; 27.83 %, n = 32/115 English). BERTScore metrics were significantly higher for German, while ROUGE metrics showed no significant differences between languages.

CONCLUSION: This study demonstrates the potential of GPT-4o for multilingual transcription of radiology reports, effectively handling both English and German with minimal errors and high semantic understanding. Future research should compare GPT-4o with current radiology dictation tools, assessing performance, cost-effectiveness, and multilingual capabilities across diverse speaker populations.

来源:Eur J Radiol,查看原文。版权归原作者所有。

引用

  1. Busch F, Prucker P, Komenda A, et al. Multilingual feasibility of GPT-4o for automated Voice-to-Text CT and MRI report transcription. Eur J Radiol. 2024. PMID: 39566177. doi:10.1016/j.ejrad.2024.111827

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。