Functional neuroimaging studies continuously show meaningful differences in metrics (e.g., localized activity, disrupted connectivity, etc.) between normal and patient populations. For any of these metrics to become a robust biomarker that can help diagnose a given medical condition, it is mandatory to prove that the observed difference is specific to the condition under study. In other words, not all differences between a certain mental disorder and healthy status can be used to diagnosis the disorder. Proving specificity of neuroimaging-based biomarkers is a difficult endeavor because of three main reasons. First, a single measure of brain function (as employed in most studies) may not be sufficient to show specificity to a certain disorder. Second, the specificity of neuroimaging biomarkers may not clearly correlate with patient populations classified solely with symptom-based diagnosis, as symptomatic diagnosis is not accurate enough to reflect differences in brain dysfunctions. Third, brain function metrics can be influenced by confounding factors such as active engagement and intentions of the patients, degrading their specificity.Our previous work, supported by the Youth Fund from NSFC, has established critical methodologies for solving these problems. The multi-subject exploratory analysis method we hav
修改评论