文献速递儿科影像神经MR

儿童脑龄预测系统在正常发育儿童中的方法学偏差

Methodological limitations of the BrainStructureAges pipeline from the volBrain platform: a brief report of the normative pediatric group.
PubMed 文献 · Pediatr Radiol · 2026-09-29

本研究对105例6~9岁正常发育儿童的脑MRI进行BrainStructureAges脑龄预测,平均预测脑龄为9.3±1.2岁,实际年龄为7.7±0.4岁,脑差值达1.6±1.2岁。95.2%的受试者呈现正向脑差值,提示该管线对儿童脑龄存在系统性高估,临床应用需谨慎解释。

要点

  • 正常儿童脑龄被高估约1.6岁,勿据单一脑龄值判断发育迟缓或超前
  • 小脑外侧面、枕叶区域偏差最明显,阅片时不宜过度解读这些区域的体积变化
  • 该管线基于成人样本训练,儿童群体中需结合临床资料综合判断

核心结论

  • BrainStructureAges管线在6~9岁正常发育儿童中存在系统性脑龄高估,平均脑差值达1.6±1.2岁
  • 95.2%的儿童在所有脑区均显示正向脑差值,提示偏差具有广泛性而非局部性

研究设计与样本

  • 回顾性分析105例健康正常发育儿童脑MRI,年龄6~9岁,均有详细病史记录
  • 所有扫描在Siemens Magnetom Amira 1.5T设备完成,采用T1加权MPRAGE序列
  • 脑龄预测使用volBrain平台上的BrainStructureAges管线,获得133个脑区的区域估计值

关键影像表现/征象

  • 偏差最大的脑区为左侧小脑外侧面(+3.3岁)和左侧枕叶梭状回(+3.3岁)
  • 小脑VI-VII小叶偏差为+2.9岁,左侧枕下回偏差为+2.9岁,右侧额极为+2.8岁
  • 各脑区以正向脑差值为主的影像模式提示广泛性体积预估不足

诊断标准与关键数值

  • 脑差值定义为实际年龄与预测脑龄之差,本组平均为1.6±1.2岁
  • 男童平均脑差值为1.55岁,女童为1.61岁,组间差异无统计学意义(t=0.27,P=0.79)
  • 95.2%的受试者在所有133个脑区中均呈现正向脑差值

临床应用与阅片要点

  • 对6~9岁儿童使用该管线进行脑龄评估时,报告应注明存在系统性高估的可能
  • 儿童小脑外侧面及枕叶脑区的脑龄偏差最大,发现这些区域异常时需考虑算法偏倚因素
  • 不宜将单一脑龄预测值作为儿童神经发育评估的独立影像依据

局限性

  • 样本年龄范围仅为6~9岁,未覆盖整个儿童期及青春期
  • 均为单一中心数据,扫描设备为1.5T,结果可能受场强和序列参数影响
  • 未与成人对照样本或不同脑龄预测管线进行交叉验证

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

BACKGROUND: Predicted brain age is a widely used neuroimaging biomarker of brain maturation. However, limited normative pediatric datasets in model training raise concerns about systematic biases when applied to children. This study evaluates the hypothesis of predicted brain age overestimation in typically developing pediatric populations.

OBJECTIVE: To analyze predicted brain age and brain delta in a large, age- and sex-matched sample of typically developing children and to test whether there is any potential systematic overestimation of predicted brain age in children using the BrainStructureAges pipeline (LaBRI, CNRS/University of Bordeaux, Bordeaux, France).

MATERIALS AND METHODS: We estimated predicted brain age from 105 brain MRI scans of healthy and typically developing children aged 6-9 years with well-documented medical histories. All scans were performed on a Magnetom Amira 1.5-T scanner (Siemens Healthineers, Erlangen, Germany), using T1-weighted MPRAGE sequences. Brain age prediction utilized the BrainStructureAges pipeline on the volBrain (ITACA Institute, Universitat Politècnica de València, Valencia, Spain, and LaBRI, CNRS/University of Bordeaux, Bordeaux, France) platform. The brain delta was defined as the difference between chronological age and predicted brain age. Regional estimates were obtained for 133 brain structures.

RESULTS: The mean predicted brain age was 9.3±1.2 years, compared to a chronological age of 7.7±0.4 years, resulting in a mean brain delta of 1.6±1.2 years. No significant sex differences were observed for chronological age, predicted brain age, or delta (boys, 1.55 years; girls, 1.61 years; t=0.27, P=0.79). Positive brain delta values were seen in 95.2% of participants across all structures, with the largest deviations in the left cerebellar exterior (+3.3 years), left occipital fusiform gyrus (+3.3 years), cerebellar lobules VI-VII (+2.9 years), left inferior occipital gyrus (+2.9 years), and right frontal pole (+2.8 years).

CONCLUSIONS: These findings suggest a systematic overestimation of predicted brain age in children, highlighting the need for caution in clinical interpretations to prevent misinterpretations in pediatric populations.

来源:Pediatr Radiol,查看原文。版权归原作者所有。

引用

  1. Korotkova, Kulikova, Lvova, et al. Methodological limitations of the BrainStructureAges pipeline from the volBrain platform: a brief report of the normative pediatric group. Pediatr Radiol. 2026. PMID: 42809125. doi:10.1007/s00247-026-06789-7

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。