Going Beyond Finding the “Lesion”: A Path for Maturation of Neuroimaging

Measurements should not be considered in isolation: interpretation of structural imaging changes should be examined side by side with functional imaging, behavioral, treatment, genetic, or other measurements on the same individuals. Interpretation of any one measure is thus strengthened by comparison with other measures.

Surpassing limited sample sizes: the emerging trend toward larger and more broadly representative sample sizes should be fostered and accelerated. This trend will help overcome frequently limited power for observing true effects and will also minimize the effects of spurious findings when assessing hundreds of thousands of voxels. This includes data sharing and pooled analyses, which come with the benefit (not cost) of capturing additional heterogeneity in the represented population (2).

Standards in reporting: despite multiple appeals to the contrary, clinical neuroimaging studies still often fail to fully report their methods, thereby impeding evaluation of potential technical confounders and undermining the increase in methodological rigor necessary for the field. We argue that the influence of technical confounders as raised by Weinberger and Radulescu (independent of issues related to the interpretation of a given measure) can be addressed given sufficient sample size and the use and documentation of well-established methods.

Replication: whether through direct replications, meta-analyses, or other forms of critical integration of previous findings, the literature can be consolidated by revealing the points of convergence (or lack thereof) across studies (3). Given the heterogeneity of psychiatric populations and the flexibility of neuroimaging experiments at the design and analytic levels, establishing robust convergence and identifying discrepancies will be crucial to the advancement of knowledge.

Causation rather than correlation: a central and often unspoken issue in the field is the inability of a correlative measure like neuroimaging to unravel causal mechanisms (2). While this limitation is true for most measures in clinical science (including behavior, genetic associations, blood measurements, etc.), the maturation of neuroimaging as a field arguably depends heavily on the link being made to causation. Thus, neuromodulation and pharmacological interventions coupled with neuroimaging may offer avenues for understanding causation while also serving as potential novel interventions.