Methods for the Analysis of Missing Data in FMRI Studies

[摘要] Functional neuroimaging has provided fundamental advances inour understanding of human brain function and is increasingly usedclinically for defining atypical function and surgical planning. Forexample, functional imaging with blood oxygenation level dependent(BOLD) contrast as a response measure is used as a clinical tool fordefining atypical development, pathology, surgical planning, andevaluating treatment outcomes. Despite years of statistical advancesin the analysis of complete whole brain data, there has been a limitedstatistical advance to address the pronounced missingness in manyfunctional imaging studies that use large discovery or small clinicalcase data. For example, functional magnetic resonance imaging(fMRI) analyses do not always include the entire brain due to imageacquisition space limitations and susceptibility artifacts (a loss andspatial distortion of signal that results from a disruption in the magneticfield). The consequence is ‘no data’ or ‘bad data’, respectively. Nodata occurs when the image acquisition doesn’t cover the whole headwhich leads to no values. In addition to susceptibility artifacts, baddata can occur across the brain because of motion artifacts. Becausestatistic maps with applied effect size or significance thresholds do nottypically include information about which voxels were omitted fromanalyses, missing data can result in Type II errors for regions that werenot tested. Missing data in fMRI studies can therefore undermine thebenefits provided by high quality imaging technology used to generatedata testing predictions about brain function.