文献速递神经MR

基于人工智能的多发性硬化纵向MRI评估:临床实践中的优势与不足

AI-based assessment of longitudinal multiple sclerosis MRI: Strengths and weaknesses in clinical practice.
PubMed 文献 · Eur J Radiol · 2025-01-18

该研究回顾性分析了2023年一家大学医院的多发性硬化随访脑MRI,按初诊与随访是否使用同一台MRI机器分为两组,用AI软件mdbrain分析新发和增大病灶,并与临床放射报告及金标准判读比较。AI软件在两组中均表现出高敏感度和高阴性预测值,但阳性预测值低,假阳性病灶的原因在两组间差异明显。

要点

  • AI软件对随访MRI中新发或增大病灶的敏感度在同机随访组为1、异机随访组为0.786。
  • 阴性预测值在同机随访组为1、异机随访组为0.954,有助于排除病灶进展。
  • 阳性预测值低(同机组0.444、异机组0.177),主要因假阳性新发或增大病灶。

核心结论

  • 对于多发性硬化患者的随访MRI评估,基于AI的影像分析在临床常规中可能有益,尤其是因其非常高的阴性预测值。

研究设计与样本

  • 研究筛选了一家大学医院2023年所有多发性硬化患者的随访脑MRI,分为「初诊与随访在同一台MRI」和「初诊与随访在不同MRI」两组,用AI软件mdbrain分析,并与临床放射报告及金标准判读比较。

诊断标准与关键数值

  • 在同机随访组,AI软件对新发或增大病灶的敏感度为1、特异度为0.74、阴性预测值为1、阳性预测值为0.444;在异机随访组,敏感度为0.786、特异度为0.549、阴性预测值为0.954、阳性预测值为0.177。

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

OBJECTIVES: In Multiple Sclerosis (MS) cerebral MRI is essential for disease and treatment monitoring. For this purpose, software solutions are available to support the radiologist with image interpretation and reporting of follow up imaging. Aim of this study was to evaluate an AI based software for longitudinal lesion detection with clinical data and to determine the influence of different MRI machines in such setting.

METHODS: The database of a university hospital was screened for all follow up MRI of MS patients performed in 2023. The examinations were categorized in "initial and follow up imaging at the same MRI" or "initial and follow up imaging at different MRI". The examinations were analysed with the AI based software mdbrain. The results concerning new and enlarging lesions were compared with the clinical radiologic report and with a gold standard reading.

RESULTS: 101 MRIs were performed at the same MRI machine and 130 at different scanners. The AI based software had a high sensitivity (1 and 0.786) and an acceptable specificity (0.74 and 0.549) concerning new or enlarging lesions in both settings. The negative predictive value was high (1 and 0.954), whereas the positive predictive value was low due to false positive new or enlarging lesions (0.444 and 0.177). The reasons for false positive lesions differed markedly in both groups.

CONCLUSION: For the evaluation of follow up MR images of MS patients, an AI-based imaging analysis can be beneficial in clinical routine, especially due to a very high negative predictive value.

来源:Eur J Radiol,查看原文。版权归原作者所有。

引用

  1. Peters S, Schmill L, Gless CA, et al. AI-based assessment of longitudinal multiple sclerosis MRI: Strengths and weaknesses in clinical practice. Eur J Radiol. 2025. PMID: 39855145. doi:10.1016/j.ejrad.2025.111941

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。