Magnetic resonance imaging (MRI) is a valuable and increasingly used method for studying brain anatomy as it allows large-scale, high-quality in vivo analyses. However, some artifacts might influence the digital results, and thus require cautious interpretation. In a recent review, these issues are addressed along with possible solutions. First, we need to keep in mind that the images acquired are not mere photographs of the brain, but reflect some biophysical properties of the tissues, by measuring the radio-frequency signals emitted by hydrogen atoms (present in water and fat) after being excited by magnetic waves. Thus, MRI is an indirect analysis of the brain anatomy and depends greatly on specific tissue properties. Second, researchers can choose from a variety of methods, depending on the aim of the survey. Macrostructure, i.e. the size and shape across voxels, can be studied through manual volumetry or automatic segmentation, voxel- or deformation-based morphometry, surface- based algorithms, or diffusion tractography. Microstructure, i.e. within-voxel contents, is usually analyzed through diffusion MRI, but also magnetization transfer imaging, or quantitative susceptibility mapping.
When making inferences on the biological significance of the outputs, the researcher must account for the possible digital artifacts. These can occur both during image acquisition and processing and can be subject-related and methodological-related. A common problem is subject motion, which might contaminate or influence the results, as the amount of motion varies with other factors influencing brain changes (age, sex, and disease status), or can even correlate with a specific effect being studied. For instance, motion induces gray matter reduction, which might be perceived as brain atrophy. Subject motion is unavoidable, but its influence can be reduced by using a motion detector during acquisition, or by estimating the amount of motion allowing statistical adjustments, also useful to detect outliers. The difficulty in manipulating the magnetic and radio-frequency fields might also introduce deformation. The main magnetic field should be spatially uniform, but it is dispersed by brain tissue while concentrated by air. This can be partially compensated by applying additional fields. The radiofrequency field is not homogeneous, which affects MRI contrast and intensity. Combining multiple transmit coils might help reduce this caveat, although the contribution and sensitivity of each coil must be taken into account when processing the image.
A particular case that can affect estimates of cortical volume and thickness is the difficulty in discriminating the dura and gray matter due to the similar intensity and anatomical proximity. In this case, MRI parameters can be manipulated in order to increase the contrast between these tissues, without reducing the contrast between gray and white matter. Individual variability in folding patterns is a further major issue in voxel-based morphometry studies because it complicates the mapping of correspondences between subjects. Registration might be enhanced by analyzing regions with larger variation to find possible anatomical alterations, aligning cortical folding patterns to locate corresponding areas, and mapping sulcal changes to improve sulci identification. Finally, researchers should continuously keep track of the constant advances and innovations in the field. The authors conclude acknowledging the importance of structural MRI when coupled with other biological information, like genetic expression (Allen Brain Atlas), cytoarchitecture (JuBrain), and cognitive associations (Neurosynth).