文献速递骨肌神经MR

人工智能驱动的腰骶神经根异常三维MRI:准确性、发生率与临床效用

Artificial intelligence-driven 3D MRI of lumbosacral nerve root anomalies: accuracy, incidence, and clinical utility.
PubMed 文献 · Neuroradiology · 2025-03-01

本研究利用人工智能软件从MRI自动生成腰骶神经根三维图像,在1500例患者中评估神经根异常的发生率与形态,并与传统人工方法对比图像准确性;结果显示神经根异常发生率为3.5%,AI自动生成图像与人工方法在1493/1500例(99.5%)中匹配,未匹配的7例均为神经根异常且被AI软件检出。

要点

  • 腰骶神经根异常在1500例中发生率为3.5%(53例,共58条神经根异常)。
  • L5-S1水平最常见,占60.3%(35条);Neidre-MacNab 1型占81.0%(47条)。
  • AI自动生成的三维神经根图像与人工方法在99.5%(1493/1500)病例中匹配。

核心结论

  • MR神经根三维成像可三维可视化并理解神经根形态(包括神经根异常),AI软件使三维神经根成像简便且精确,对脊柱手术术前评估有很大帮助。

研究设计与样本

  • 在1500例患者的三维图像中评估神经根异常的发生率与形态,并将AI软件自动生成的图像与传统方法人工创建的图像进行比较以评估准确性。

关键影像表现/征象

  • 1500例中53例(3.5%)存在神经根异常,共58条神经根异常。
  • 按脊柱水平,L5-S1水平35条,最常见(60.3%)。
  • 按形态,47条(81.0%)为Neidre-MacNab分类1型。

诊断标准与关键数值

  • AI自动生成图像与人工方法在1500例中1493例(99.5%)匹配,剩余7例均存在神经根异常,且被AI软件检出为异常。

临床应用与阅片要点

  • 作者指出,AI软件可实现简便且精确的三维神经根成像,对脊柱手术术前评估有很大帮助。

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: Lumbosacral nerve root anomalies are relatively rare but can be a risk factor for intraoperative nerve injury. However, it is often difficult to evaluate them with preoperative imaging. We developed a software that automatically generates three-dimensional (3D) nerve root images from magnetic resonance (MR) imaging using artificial intelligence (AI). This study aims to evaluate the accuracy and utility of this modality in clinical practice by conducting an epidemiological study of nerve root anomalies.

METHODS: The incidence and morphology of nerve root anomalies were evaluated in the 3D images of 1,500 patients. The accuracy of the images was evaluated by comparing the images generated automatically using this AI software with those created manually by conventional methods.

RESULTS: Of 1,500 cases, 53 (3.5%) had nerve root anomalies with total of 58 nerve root anomalies. With respect to the spinal level, 35 nerve root anomalies were found in the L5-S1 level, the most common (60.3%). As for morphology, 47 nerve roots (81.0%) were of the Neidre-MacNab classification Type 1. The images matched in 1,493 out of 1,500 cases (99.5%) between the two methods, and the remaining 7 cases all had nerve root abnormalities, which were detected as abnormal by the AI software.

CONCLUSION: The MR nerve root 3D imaging provided a 3D visualization and understanding of nerve root morphology, including nerve root anomalies. The AI software enables easy and precise 3D nerve root imaging, which greatly aids in the preoperative evaluation for spinal surgery.

来源:Neuroradiology,查看原文。版权归原作者所有。

引用

  1. Ukeba D, Nagahama K, Yamada K, et al. Artificial intelligence-driven 3D MRI of lumbosacral nerve root anomalies: accuracy, incidence, and clinical utility. Neuroradiology. 2025. PMID: 40021500. doi:10.1007/s00234-025-03574-5

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。