基于结构MRI的机器学习识别弗里德赖希共济失调的生物学亚型
这项多中心二次分析纳入565名参与者(275例弗里德赖希共济失调患者和290名健康对照),使用SuStaIn算法分析结构MRI生物标志物。研究识别出三种结构MRI亚型:经典型(66.5%)、早期大脑型(25.8%)和早期小脑型(7.64%),各亚型具有不同的脑退化模式和临床严重度关联。
要点
- SuStaIn算法基于结构MRI将弗里德赖希共济失调分为经典型、早期大脑型和早期小脑型三种亚型。
- 经典型占比最高(66.5%,183/275),表现为从脑干到小脑皮层再到大脑的上升梯度损伤。
- 早期大脑型(25.8%,71/275)大脑萎缩先于小脑皮层受累;早期小脑型(7.64%,21/275)小脑小叶萎缩先于上脑干或大脑受累。
核心结论
- 使用SuStaIn算法识别出三种不同的弗里德赖希共济失调结构MRI亚型,各亚型具有不同的脑退化模式并与临床严重度存在关联。
研究设计与样本
- 多中心二次分析,纳入来自ENIGMA-Ataxia工作组的565名参与者(275例弗里德赖希共济失调患者和290名健康对照),使用结构MRI生物标志物和SuStaIn算法进行亚型分层。
关键影像表现/征象
- 经典型亚型(66.5%,183/275)表现为从脑干到小脑皮层再到大脑的上升梯度损伤。
- 早期大脑型亚型(25.8%,71/275)大脑萎缩先于小脑皮层受累。
- 早期小脑型亚型(7.64%,21/275)小脑小叶萎缩先于上脑干或大脑受累。
诊断标准与关键数值
- SuStaIn算法基于结构MRI生物标志物识别亚型并分期;更晚的MRI-based SuStaIn分期与更长的症状持续时间相关(非标准化系数B=0.422,标准误=0.065,P<.001)。
- 更晚的MRI-based SuStaIn分期与更大的症状严重度相关(B=1.404,标准误=0.201,P<.001)。
- 早期大脑型亚型的生物学分期进展与临床变量的映射较其他亚型更弱(交互项早期大脑型亚型×分期:B=-0.925,标准误=0.410,P=.02)。
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
Background Friedreich ataxia (FRDA) is an inherited, progressive neurodegenerative disease. Interindividual heterogeneity in the rate and phenotypic profile of disease progression indicates a biologic variability in the pattern and spatial evolution of underlying changes, but the occurrence of possible FRDA subgroups, which could aid in clinical trial design and treatment, are still unknown. Purpose To obtain a structural MRI-based stratification of participants with FRDA using the Subtype and Stage Inference (SuStaIn) algorithm and determine whether these subgroups are biologically meaningful and clinically relevant. Materials and Methods This multicenter secondary analysis of prospectively acquired data included structural MRI and clinical-demographic data from participants from the ENIGMA-Ataxia working group. MRI biomarkers were analyzed using the SuStaIn algorithm to identify subgroups with distinct patterns and disease stages. The clinical and genetic relevance of these subgroups were assessed within a linear model framework. Results This study included 565 participants (mean age, 32 years ± 13.1 [SD]; 286 women; 275 participants with FRDA and 290 healthy controls). SuStaIn identified three subtypes: (a) a classical subtype (66.5% [183 of 275 participants]), characterized by an ascending gradient of damage from brainstem to cerebellar cortex to cerebrum; (b) an early cerebral subtype (25.8% [71 of 275 participants]) with cerebral atrophy preceding the involvement of cerebellar cortex; and (c) and an early cerebellar subtype (7.64% [21 of 275 participants]) showing cerebellar lobule atrophy before upper brainstem or cerebral involvement. More advanced disease stages (MRI-based SuStaIn stages) correlated with greater symptom duration (unstandardized coefficient B = 0.422, standard error = 0.065, P < .001) and severity (B = 1.404, standard error = 0.201, P < .001), and these relationships were moderated by subtype, with biologic stage progression in the early cerebral subtype mapping less strongly to clinical variables relative to the others (interaction term early cerebral subtype × stage: B = -0.925, standard error = 0.410, P = .02). Conclusion Using the SuStaIn algorithm, three distinct structural MRI-based subtypes of FRDA were identified, with different patterns of brain degeneration and associations with clinical severity. © RSNA, 2026 Supplemental material is available for this article.
引用
- Pontillo G, Penna S, Arrigoni F, et al. Identification of Biological Subtypes of Friedreich Ataxia with Structural MRI-based Machine Learning. Radiology. 2026. PMID: 41805414. doi:10.1148/radiol.251386
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。