Avoiding performance and task confounds: multimodal investigation of brain reorganization after stroke rehabilitation.
Avoiding performance and task confounds: multimodal investigation of brain reorganization after stroke rehabilitation.
复制标题
避免表现和任务混淆：中风康复后大脑重组的多模式研究。
DOI：
10.1016/j.expneurol.2006.12.026
复制
发表时间：
2007
影响因子：
5.3
通讯作者：
Krakauer,JohnW
中科院分区：
文献类型：
--
作者：
Krakauer,JohnW
尽管在预防和急性治疗方面取得了重大进展，但中风仍是全球残疾的主要原因。因此，迫切需要了解自发和康复诱导恢复的神经机制。在过去的15年中，功能性脑成像已被广泛用于研究全脑水平的恢复相关变化（Calautti和Baron，2003; Cramer等人，1997; Krakauer，2004;马歇尔等人，2000年; Ward等人，2003 a，B）。然而，很明显，当单独使用功能成像来研究中风后的脑重组时，已经接近僵局。这里简要描述其原因，因为它们与运动恢复有关。首先，运动恢复的定义一直令人惊讶地强调不足，即使很明显，激活的区域或模式将取决于扫描仪中的运动任务和扫描仪外的行为测量选择作为图像分析中的预测因素。其次，恢复的措施往往是粗糙的，对补偿性改善与真正的恢复或功能替代之间的差异不敏感。这是功能成像的一个真实的问题，因为代偿性调整也常常导致激活变化，即使它们与真正的恢复无关。例如，使用更多的近端肢体肌肉来辅助远端控制可能会导致对侧病变激活，因为近端肌肉具有更多的双侧皮质代表，但当然这种新的激活不会表明中风后的重组。如果严格地理解为由于相同的神经计算（尽管是在新的解剖部位进行的）而恢复到相同的病前行为，那么真正的恢复可能永远不会发生，当考虑到这一点时，情况变得更糟。这意味着必须仔细选择任务和测量方法，以便研究人员知道他们将激活变化归因于什么。第三，功能成像是一种相关技术，充其量只能证明大脑区域对行为的必要性，而不是充分性。识别新的激活区域本身并不能告诉研究人员该区域在计算中正在做什么。一个隐含的但不合理的假设，在文献中很常见，是一个新的区域只涉及恢复，如果它有助于运动执行。或者，在研究A中，一个区域执行x功能，因为在研究B中，当x功能直接与任务y一起测试时，它激活了相同的区域，这是参与消极推理（Poldrack，2006），最好避免。最后但并非最不重要的是，医院中MR扫描仪的使用增加，成像分析软件包的用户友好性，以及图像本身的纯粹诱惑力，导致人们容易忽视困扰功能成像的混淆，特别是功能恢复的问题。其中最有问题的是表现混淆，其他包括控制受试者注意力、努力和任务难度感知的挑战。表现混淆可以理解为：中风患者受损，因此他们将与对照受试者不同地执行给定的运动任务。这种表现差异本身就可能导致激活的差异，并混淆了更有趣的解释，即不同的激活代表了维持运动输出的重组。这种混淆也是纵向研究的一个破坏因素，因为根据定义，随着患者的康复，他们的表现也会改善。唯一令人信服的恢复相关的结果将是，如果...
Stroke is the leading cause of disability worldwide despite significant advances in prevention and acute treatment. Thus there is an urgent need to understand the neural mechanisms of both spontaneous and rehabilitation-induced recovery. In the past 15 years, functional brain imaging has been used extensively to investigate recovery-related changes at the whole brain level (Calautti and Baron, 2003; Cramer et al., 1997; Krakauer, 2004; Marshall et al., 2000; Ward et al., 2003a, b). However, it has become apparent that functional imaging, when used alone to investigate post-stroke brain reorganization, has reached a near impasse. The reasons for this are briefly delineated here as they pertain to motor recovery. First, the definition of motor recovery has been surprisingly underemphasized even though it is obvious that the regions or patterns of activation identified will depend on the motor task in the scanner and on the out of scanner behavioral measures chosen as predictors in the image analysis. Second, measures of recovery are often crude and insensitive to the difference between compensatory improvements versus true recovery, or vicariation, of function. This is a real problem for functional imaging because compensatory adjustments will also often lead to activation changes even though they have nothing to do with true recovery. For example, use of more proximal limb muscles to aid distal control might lead to contralesional activation, as proximal muscles have more bilateral cortical representation, but of course this novel activation would not indicate reorganization after stroke. The situation becomes even worse when one considers that true recovery may in fact never occur, if strictly understood to mean a return to identical pre-morbid behavior due to identical neural computations, albeit performed at a new anatomical site. The implication is that tasks and measures have to be carefully chosen so that investigators know what they are attributing activation changes to. Third, functional imaging is a correlative technique and at best can prove necessity but not sufficiency of a brain area to behavior. Identification of a novel area of activation does not in itself tell the investigator what the area is doing computationally. An implicit but unjustified assumption, common in the literature, is that a novel area is only involved in recovery if it contributes to motor execution. Alternatively, to conclude that an area performs function x in study A because when function x was directly tested with task y in study B it activated the same area is to engage in negative inference (Poldrack, 2006) and is best avoided. Last but not the least, the increased access to MR scanners in hospitals, the user friendliness of imaging analysis software packages, and the sheer seductiveness of the images themselves have led to a susceptibility to lose sight of the confounds that plague functional imaging in general, and issues of restoration of function in particular. The most problematic of these is the performance confound, others include the challenge of controlling subject attention, effort, and perception of task difficulty.The performance confound can be understood as follows: patients with stroke are impaired and therefore they will perform a given motor task differently from control subjects. This performance difference alone could lead to a difference in activation and confound the more interesting interpretation, which is that the different activation represents reorganization to maintain motor output. This confound is also a spoiler for longitudinal studies because as patients recover, by definition, their performance also improves. The only compelling recovery-related result would be if …