Unpacking the Biases That Shape the Apparent Foci in the Meta-analysis of Voxel-Based Neuroimaging Studies.
Unpacking the Biases That Shape the Apparent Foci in the Meta-analysis of Voxel-Based Neuroimaging Studies.
复制标题
在基于体素的神经影像研究的荟萃分析中揭示形成明显病灶的偏差。
DOI：
10.1016/j.biopsych.2022.06.021
复制
发表时间：
2022
影响因子：
10.6
通讯作者：
Chung,Yoonho
中科院分区：
文献类型：
--
作者：
Chung,Yoonho
虽然最新的诊断分类系统继续将精神分裂症（SCZ）和双相情感障碍（BD）归类为不同的疾病实体，但这些疾病的二分概念多年来逐渐减弱。从家族到分子研究的证据表明，这两种疾病之间的遗传病因基本上重叠（1）。这促使临床神经影像学研究人员将SCZ和BD纳入研究样本中，以阐明与基础生物系统相关的跨诊断临床特征（2）。有大量的神经影像学研究报告，通过将SCZ和BD患者与正常对照受试者进行比较，发现了重叠和区域性脑异常（3）。然而，由于疾病慢性化和药物暴露的混杂效应，这种研究设计对揭示大脑中表达的特定疾病效应提出了挑战。克服这一挑战的一种策略是研究未受影响的生物学家庭成员，他们通常没有这种混淆，因为他们预计会遗传与该疾病的遗传易感性相关的神经风险决定因素。到目前为止，大量的研究已经应用基于体素的分析比较了SCZ患者的未受影响的亲属（SCZ-REL）和BD患者的亲属（BD-REL）与健康对照受试者，并发现亲属显示出与患者组相似的结构和功能性大脑改变的模式，但程度较低。然而，研究结果是异质性的，并且协调大量的高维全脑地图是一项具有挑战性的任务。(4)已经做了一个令人印象深刻的工作，把一个全面的荟萃分析结构和任务为基础的功能性磁共振成像（fMRI）研究的个人谁是在家族风险SCZ和BD。值得注意的是，使用激活似然估计方法进行了基于坐标的荟萃分析，以汇编全脑分析结果（5）。本质上，该方法将研究中立体定向病灶的坐标（与代表最大病例对照平均差异的分布的统计阈值“峰值”相对应的聚集体素）制成表格，以定量推断在体素水平发现组间差异的可能性。作为研究的一部分，将不同任务中基于体素的形态测量学和功能磁共振成像研究的焦点汇集在一起，以确定在成像方式中显示最常见的组差异的大脑区域。使用所描述的方法，Cattarinussi et al. (4)提供了显示病灶的体素统计图，这些病灶可能反映了与病理生物学相关的大脑变化。然而，对这些明显病灶的解释应谨慎对待，因为在个体研究中应用的基于体素的统计方法（即，质量单变量分析）在揭示特定疾病影响方面具有已知的局限性（6）。
Although the latest diagnostic classification system continues to categorize schizophrenia (SCZ) and bipolar disorder (BD) as distinct disease entities, the dichotomic conceptualization of these illnesses has progressively weakened over the years. Converging evidence from family to molecular studies shows that the genetic etiology between these two disorders substantially overlaps (1). This has prompted clinical neuroimaging researchers to include both SCZ and BD in the study sample to elucidate transdiagnostic clinical features that are linked to the underlying biological systems (2). There are a vast number of neuroimaging studies reporting overlapping and district brain abnormalities between patients with SCZ and BD by comparing them to normal control subjects (3). However, such study designs pose a challenge for uncovering specific disease effects expressed in the brain because of the confounding effects of illness chronicity and exposure to medications. One strategy to overcome this challenge is by studying unaffected biological family members who generally are free of such confounds, as they are expected to inherit neural risk determinants that are associated with the genetic liability of that disorder. To date, a substantial number of studies have applied voxel-based analysis comparing unaffected relatives of patients with SCZ (SCZ-REL) and relatives of patients with BD (BD-REL) to healthy control subjects and found that the relatives show a pattern of structural and functional brain alterations that is similar to the patient group, but to a lesser degree. However, the findings are heterogeneous, and reconciling a large quantity of whole-brain maps that are high-dimensional in nature is a challenging task.In the current issue of Biological Psychiatry, Cattarinussi et al.(4) have done an impressive job putting together a comprehensive meta-analysis of structural and task-based functional magnetic resonance imaging (fMRI) studies of individuals who are at familial risk for SCZ and BD. Notably, coordinate-based meta-analysis using the activation likelihood estimation method was performed to compile whole-brain analysis results (5). In essence, this method tabulates the coordinates of stereotactic foci (clustered voxels corresponding to statistically thresholded “peaks” of a distribution representing maximal case-control mean differences) across studies to make quantitative inferences about the likelihood of finding group differences at the voxel level. As part of the study, foci from voxel-based morphometry and fMRI studies across different tasks were pooled together to identify brain regions that showed the most frequent group differences across imaging modalities. Using the method described, Cattarinussi et al.(4) provided voxelwise statistical maps showing foci that may be reflecting changes in the brain related to the pathobiology. However, interpretations of those apparent foci should be approached with caution as the voxel-based statistical methods (ie, mass univariate analysis) applied in individual studies have known limitations in revealing specific disease effects (6).