文献速递神经儿科影像MR

自动化定量评估脑MRI可能更准确地区分早产成人

Automated quantitative evaluation of brain MRI may be more accurate for discriminating preterm born adults.
PubMed 文献 · Eur Radiol · 2019-03-22

研究纳入67名成人(27名足月对照,40名极早产/极低出生体重者),比较自动分割、手工测量与视觉评级对脑结构异常的诊断准确性。自动分割正确分类84%的病例,视觉评级为63%,两者结合达87%。极早产/极低出生体重成人中侧脑室扩大变形、胼胝体变薄及白质「不干净」表现常见,但视觉评估诊断准确性低。

要点

  • 极早产/极低出生体重成人脑MRI常见侧脑室扩大变形、胼胝体变薄和白质「不干净」。
  • 视觉评级对上述异常的诊断准确性低(正确分类63%),自动分割定量更准确(84%)。
  • 临床阅片见上述征象时,作者建议先询问早产史再考虑其他诊断。

核心结论

  • 极早产/极低出生体重成人中侧脑室扩大变形、胼胝体变薄和白质「不干净」常见;视觉评估诊断准确性低,自动体积定量更准确但耗时;作者建议对具有这些改变者进一步诊断前询问早产史。

研究设计与样本

  • 2011-2013年采集的3-T MRI,67名成人(27名足月对照,平均年龄26.4岁,8名女性;40名极早产/极低出生体重者,平均年龄26.6岁,16名女性),比较自动分割、手工胼胝体测量及侧脑室和白质视觉评级,采用t检验、logistic回归和ROC曲线分析。

关键影像表现/征象

  • 极早产/极低出生体重成人侧脑室体积增大(p < 0.01)。
  • 极早产/极低出生体重成人胼胝体后部体积降低(p < 0.01)。
  • 极早产/极低出生体重成人深部灰质体积降低(p < 0.01)。
  • 视觉评级和手工测量显示极早产/极低出生体重成人胼胝体更薄(p = 0.04)、侧脑室变形(p = 0.03),白质「不干净」有增多趋势(p = 0.06)。

诊断标准与关键数值

  • 自动分割正确分类84%的病例,视觉评级正确分类63%的病例。
  • 自动/手工测量联合视觉评级正确分类87%的病例。
  • 逐步logistic回归识别出三个独立特征正确分类81%的病例:侧脑室体积、深部灰质体积和白质表现。

临床应用与阅片要点

  • 在临床实践中,若MRI上可见侧脑室大且变形、胼胝体小且薄以及白质「不干净」,作者建议在考虑其他诊断前询问早产史。
  • 视觉发现虽常见但准确性低;加入侧脑室和深部灰质核团的自动分割可提高诊断准确性。
  • 自动体积定量更准确但耗时。

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

OBJECTIVE: To investigate the structural brain abnormalities and their diagnostic accuracy through qualitative and quantitative analysis in term born and very preterm birth or with very low birth weight (VP/VLBW) adults.

METHODS: We analyzed 3-T MRIs acquired in 2011-2013 from 67 adults (27 term born controls, mean age 26.4 years, 8 females; 40 VP/VLBWs, mean age 26.6 years, 16 females). We compared automatic segmentations of the white matter, deep gray matter and cortical gray matter, manual corpus callosum measurements and visual ratings of the ventricles and white matter with t tests, logistic regression, and receiver operator characteristic (ROC) curves.

RESULTS: Automatic segmentation correctly classified 84% of cases; visual ratings correctly classified 63%. Quantitative volumetry based on automatic segmentation revealed higher ventricular volume, lower posterior corpus callosum, and deep gray matter volumes in VP/VLBW subjects compared to controls (p < 0.01). Visual rating and manual measurement revealed a thinner corpus callosum in VP/VLBW adults (p = 0.04) and deformed lateral ventricles (p = 0.03) and tendency towards more "dirty" white matter (p = 0.06). Automatic/manual measures combined with visual ratings correctly classified 87% of cases. Stepwise logistic regression identified three independent features that correctly classify 81% of cases: ventricular volume, deep gray matter volume, and white matter aspect.

CONCLUSION: Enlarged and deformed lateral ventricles, thinner corpus callosum, and "dirty" white matter are prevalent in preterm born adults. Their visual evaluation has low diagnostic accuracy. Automatic volume quantification is more accurate but time consuming. It may be useful to ask for prematurity before initiating further diagnostics in subjects with these alterations.

KEY POINTS: • Our study confirms prior reports showing that structural brain abnormalities related to preterm birth persist into adulthood. • In the clinical practice, if large and deformed lateral ventricles, small and thin corpus callosum, and "dirty" white matter are visible on MRI, ask for prematurity before considering other diagnoses. • Although prevalent, visual findings have low accuracy; adding automatic segmentation of lateral ventricles and deep gray matter nuclei improves the diagnostic accuracy.

来源:Eur Radiol,查看原文。版权归原作者所有。

引用

  1. Jurcoane A, Daamen M, Keil VC, et al. Automated quantitative evaluation of brain MRI may be more accurate for discriminating preterm born adults. Eur Radiol. 2019. PMID: 30903339. doi:10.1007/s00330-019-06099-7

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。