METHODS: We included individuals with and without pharmaco-resistant TLE who underwent multimodal 3T MRI, including arterial spin labelling, structural, and diffusion-weighted imaging. Using surface-based MRI mapping, we generated individualized cortico-subcortical profiles of perfusion, morphology, and microstructure. Linear models compared regional CBF in patients with controls and related alterations to morphological and microstructural metrics. We further probed interregional vascular networks in TLE, using graph theoretical CBF covariance analysis. The effects of disease duration were explored to better understand the progressive changes in perfusion. We assessed the utility of perfusion in separating patients with TLE from controls using supervised machine learning.
RESULTS: Compared with control participants (n = 38; mean ± SD age 34.8 ± 9.3 years; 20 females), patients with TLE (n = 24; mean ± SD age 35.8 ± 10.6 years; 12 females) showed widespread CBF reductions predominantly in fronto-temporal regions (Cohen d -0.69, 95% CI -1.21 to -0.16), consistent in a subgroup of patients who remained seizure-free after surgical resection of the seizure focus. Parallel structural profiling and network-based models showed that cerebral hypoperfusion may be partially constrained by gray and white matter changes (8.11% reduction in Cohen d) and topologically segregated from whole-brain perfusion networks (area under the curve -0.17, p < 0.05). Negative effects of progressive disease duration further targeted regional CBF profiles in patients (r = -0.54, 95% CI -0.77 to -0.16). Perfusion-derived classifiers discriminated patients from controls with high accuracy (71% [70%-82%]). Findings were robust when controlling for several methodological confounds.
CONCLUSIONS: Our multimodal findings provide insights into vascular contributions to TLE pathophysiology affecting and extending beyond mesiotemporal structures and highlight their clinical potential in epilepsy diagnosis. As our work was cross-sectional and based on a single site, it motivates future longitudinal studies to confirm progressive effects, ideally in a multicentric setting.
方法:我们纳入了接受多模态3TMRI的有或没有药物耐药TLE的个体,包括动脉自旋标记,结构,和弥散加权成像。使用基于表面的MRI映射,我们生成了个体化的皮质-皮质下灌注图,形态学,和微观结构。线性模型比较了对照组患者的区域CBF以及形态学和微结构指标的相关改变。我们进一步研究了TLE中的区域间血管网络,采用图论CBF协方差分析。探讨了疾病持续时间的影响,以更好地了解灌注的进行性变化。我们使用监督机器学习评估了灌注在将TLE患者与对照组分离中的效用。
结果:与对照组参与者相比(n=38;平均±SD年龄34.8±9.3岁;20名女性),TLE患者(n=24;平均±SD年龄35.8±10.6岁;12名女性)表现出广泛的CBF降低,主要在额颞区(Cohend-0.69,95%CI-1.21至-0.16),在手术切除癫痫灶后仍无癫痫发作的患者亚组中一致.并行的结构分析和基于网络的模型显示,脑灌注不足可能部分受到灰质和白质变化的约束(Cohend减少8.11%),并在拓扑上与全脑灌注网络隔离(曲线下面积-0.17,p<0.05)。进行性疾病持续时间的负面影响进一步针对患者的区域CBF概况(r=-0.54,95%CI-0.77至-0.16)。灌注衍生分类器以高准确度(71%[70%-82%])将患者与对照区分开来。当控制几个方法学困惑时,研究结果是稳健的。
结论:我们的多模态研究结果提供了对TLE病理生理学影响并延伸超过中颞叶结构的血管贡献的见解,并强调了其在癫痫诊断中的临床潜力。由于我们的工作是横截面的,并且基于一个站点,它激发了未来的纵向研究来确认渐进效应,理想情况下,在一个多中心的设置。