Independent Researcher.
The Australian National University.
Stud Health Technol Inform. 2024 Sep 24;318:78-83. doi: 10.3233/SHTI240895.
Machine Translation (MT) has emerged as a crucial tool in bridging language barriers. In health settings, MT is increasingly relevant due to the diversity of patient populations, the dominance of English in medical research, and the limited availability of human translation services. Improvements in MT accuracy have prompted a re-evaluation of its suitability in contexts where it was once deemed impractical. This scoping review with meta-analysis delved into the appropriateness and limitations of MT in health, including in medical education, literature translation, and patient-provider communication. A keyword search in PubMed, PubMed Central, and IEEE Xplore produced peer-reviewed literature that focused on MT in a health context, published from 2018 to 2023. Analysis and mapping of full-text articles revealed 33 studies among 2,589 returned abstracts, indicating that MT is still unsuitable for direct use in patient interactions, due to clinical risks linked to insufficient accuracy. However, MT was showing promise further away from patients, for translation of medical articles, terminology, and educational content. Further research in improving MT performance in these contexts, coverage of under-studied languages, and study of the existing usages of MT are recommended.
机器翻译（MT）已成为弥合语言障碍的重要工具。在医疗环境中，由于患者群体的多样性、医学研究中英语的主导地位以及人工翻译服务的有限可用性，MT 的相关性日益增强。MT 准确性的提高促使人们重新评估其在曾经被认为不切实际的情况下的适用性。这项范围综述与荟萃分析深入探讨了 MT 在健康领域的适宜性和局限性，包括医学教育、文献翻译和医患沟通。在 PubMed、PubMed Central 和 IEEE Xplore 中进行的关键字搜索产生了同行评议的文献，这些文献重点关注 2018 年至 2023 年期间健康背景下的 MT。对全文文章的分析和映射揭示了在 2589 篇返回摘要中有 33 项研究，这表明由于与准确性不足相关的临床风险，MT 仍不适合直接用于患者交互。然而，MT 在远离患者的情况下，例如医学文章、术语和教育内容的翻译，显示出了一定的前景。建议进一步研究如何提高 MT 在这些情况下的性能、涵盖研究不足的语言，并研究 MT 的现有应用。
