基于动态低频振幅机器学习对无先兆偏头痛相关睡眠障碍的机制探究

Exploration of the Mechanistic of Sleep Disorders in Migraine Without Aura:A Machine Learning Analysis of dALFF

目的:采用动态低频振幅技术(dALFF)结合监督学习模型探讨无先兆偏头痛(患者睡眠障碍)的神经影像机制。方法:收集52名确诊为无先兆偏头痛的患者,根据匹兹堡睡眠质量指数将其分为偏头痛伴睡眠障碍组(MwSD)与偏头痛不伴睡眠障碍组(MwoSD),并同期招募40名年龄与性别相匹配的健康志愿者作为对照组。使用3.0 T MRI采集其T1WI和BOLD-fMRI图像,通过随机森林(RF)、逻辑回归(LR)、支持向量机(SVM)机器学习算法对其dALFF进行分析。结果:单因素方差分析(one-way ANOVA)显示3组受试者在左侧丘脑、右侧枕中回、左侧楔前叶dALFF有差异;两独立样本t检验显示MwSD组较MwoSD组左内侧和旁扣带回、左侧额中回dALFF值升高,右侧颞上回dALFF值下降;无监督学习模型中SVM性能较好,权重绝对值排名前三的脑区右侧枕中回、左侧楔前叶、左内侧和旁扣带回,均与视觉通路相关。结论:视觉通路相关脑区dALFF值可作为预测MwSD患者的神经影像学标志。

Purpose:To explore the sleep disorders in patients with migraine without aura(MwoA)using dynamic amplitude of low-frequency fluctuations(dALFF)technique combined with supervised learning models.Methods:Fifty-two confirmed patients diagnosed with MwoA were recruited.According to the Pittsburgh sleep quality index,they were divided into migraine with sleep disorder(MwSD)group and migraine without sleep disorder(MwoSD)group,with 40 healthy volunteers of matching age and gender as a control group simultaneously.T1WI and BOLD-fMRI images were collected using 3.0 T MRI.The dALFF was analyzed using machine learning algorithms including random forest(RF),logistic regression(LR),and support vector machine(SVM).Results:Oneway ANOVA analysis showed differences in dALFF in the left thalamus,right posterior cingulate gyrus,and left precuneus among the three groups.Independent sample t-tests revealed higher dALFF in the left medial and lateral occipital gyrus and left superior frontal gyrus and lower dALFF in the right superior temporal gyrus in the MwSD group than those in the MwoSD group.SVM performed better in unsupervised learning model,and the absolute weight of the top three brain regions were right middle occipital gyrus,left anterior cuneus,left medial and paracingulate gyrus,all of which were related to visual pathways.Conclusion:dALFF values in brain regions related to the visual pathway may serve as neuroimaging markers for predicting sleep disorders in patients with MwoA.