In their recent Review article (A pathophysiological framework of hippocampal dysfunction in ageing and disease. Nature Rev. Neurosci. 12, 585–601 (2011))1, Small et al. present a compelling framework for differentiating hippocampal disorders based on the selective vulnerability of hippocampal subregions using recent neuroimaging findings. This framework stimulates thoughts about how (dys)function of distinct hippocampal subregions relates to disease and how it can be assessed in the future using high-resolution structural and functional MRI. The impact this pathophysiological framework has on clinical practice depends to a great extent on the availability, quality and reliability of methods to discern hippocampal subregions in vivo. In many hospitals, 3-Tesla MRI scanners have become the standard for obtaining high-resolution in vivo brain images, and the introduction of 7-Tesla MRI may lead to a revolutionary increase in image detail. Nonetheless, accurately measuring hippocampal subregions with in vivo MRI has proved to be challenging2. In this Correspondence, we illustrate how human in vivo MRI acquisition and image post-processing methods need to advance to reliably measure and differentiate hippocampal subregions.

In neuroanatomy, hippocampal subregions are discerned on the basis of transitions in cytoarchitecture3 (such as the number of cortical layers or the density of a cell layer) that can be made visible in, for example, Nissl-stained tissue (Fig. 1A). Using high-field-strength ex vivo MRI on small hippocampus samples, images that approach microscopic quality (for example, 0.1 mm isotropic voxels) can be obtained4,5. However, even then, the level of spatial detail is often too limited to reliably discern hippocampal subregions (Fig. 1B,C). This is even more challenging with in vivo 3- and 7-Tesla images. With a spatial resolution of 1.5 mm isotropic voxels in functional MRI images and more than 0.4 mm in structural images (Fig. 1D,E), the size of the smallest measuring unit (voxel) substantially exceeds the thickness of a cortical layer, resulting in an inability to reliably see cortical layers or layer transitions in such images.

Figure 1: Histology, ex vivo and in vivo MRI of the human hippocampus.
figure 1

A | 50 μm thick Nissl-stained oblique horizontal section. Ba | Same tissue as A at a similar angle, T1-weighted image at 7 Tesla (resolution: 0.25 mm isotropic voxels; acquisition time: 30 h, 15 min). The red line shows the plane of the cross section shown in Bb. Bb | Oblique coronal image. C | Coronal T2-weighted image of the Penn hippocampus atlas10 measured at 9.4 Tesla (resolution: 0.2 mm isotropic voxels; acquisition time: 62.5 h). D | Oblique coronal T2-weighted image measured at 3 Tesla in vivo; same subject as E at approximately the same level (resolution: 0.75 mm inplane; slice thickness 0.8 mm; acquisition time: 18 min). The hippocampal formation is located in the red circles. E | Oblique coronal image measured with 3.Tesla in vivo functional MRI; same subject as D (resolution: 1.5 mm in-plane; slice thickness 3.0 mm; acquisition time per volume: 2.0 s). DG, dentate gyrus; Sub, subiculum.

In current in vivo neuroimaging studies in which hippocampal subfields are discerned, either a manual or automated segmentation procedure is applied that is based on visual or calculated similarity between the obtained magnetic resonance image and a detailed anatomical atlas of the hippocampal subdivisions. However, no reliable, quantifiable relationship between macroscopic anatomical landmarks (for example, folding patterns of gyri) and the exact location of hippocampal subregions is known to exist6, despite reports that in other cortical regions, macroscopic features can be predictive for the underlying cytoarchitecture7. In healthy individuals, a probabilistic atlas may be used to determine the probability of the location of hippocampal subregions8. However, this atlas is based on healthy individuals and is likely to be invalid for determining subregions in individuals suffering from hippocampus-related pathology, as the pathology distorts the geometry of the hippocampal subregions differentially9. The Review by Small et al. provides an excellent starting point to advance knowledge about the relationship between high-resolution in vivo and ex vivo MRI and histopathological images, yet clinical relevance will only increase if methods that allow accurate localization of hippocampal subfields using in vivo MRI become available.