Is the whole brain periventricular?

Short abstract

The term periventricular is becoming meaningless and should be abandoned

The unique sensitivity of MRI to detect brain tissue alterations, especially white matter lesions (WML), has intrigued researchers ever since the introduction of this modality in the 1980s. Following early reports on the apparently meaningless nature of WML, and the naive labelling of WML as “unidentified bright objects” or UBOs,¹ a wealth of studies have examined the relevance of WML in normal and pathological ageing. This has yielded strong and converging evidence regarding risk factors for WML and, to a lesser extent, the clinical significance of WML for cognitive dysfunction. Contributing to the latter are factors such as study design (cross sectional v longitudinal), selection bias (population based v case‐control), type of cognitive assessment (dementia screening tools v detailed neuropsychology), and MRI methodology (choice of scan acquisition and analysis technique).

In the quest for elucidation of the significance of WML for cognitive functioning, the issue of localisation of WML has gained considerable attention, since it is obvious that certain locations are more relevant than others, depending on the type of function probed. In the current issue, Van den Heuvel and colleagues (see pages 149–53)² report an increase in periventricular, but not subcortical, WML in parallel with cognitive decline in a population based study on non‐demented subjects. These longitudinal data are in agreement with results from the cross sectional Rotterdam Scan Study³ to the extent that they suggest that subcortical lesions may be less relevant (for certain cognitive functions). While one can certainly formulate a plausible hypothesis to support the specific clinical impact of periventricular lesions (disrupting important descending and association fibres traversing this area), such speculations depend very much on the definition of what constitutes a “periventricular” WML.

MRI‐histopathological studies have revealed that a commonly encountered feature on scans of aged individuals, so called “caps” and “bands” of high T2 signal, reflect only mild pathology in the 5–10 mm zone directly adjacent to the ventricles.⁴ Since the associated histopathological changes (disruption of the ependymal lining, myelin pallor, and some subependymal astrogliosis) are relatively mild, these strictly periventricular changes (within 10 mm of the ventricular lining) are singled out in certain visual rating scales.⁵,⁶ Other studies, however, apply the term periventricular much more loosely: van den Heuvel et al report that approximately 80% of WML are “periventricular”, defined as any WML having contact with the ventricle; in the Rotterdam Scan Study, this percentage is around 60%.³ It is obvious that labelling such a high percentage of lesions as being periventricular deviates from the previously mentioned data on “caps” and “bands”.

When WML grow, they have a tendency to become confluent, and in the end such confluent abnormalities are increasingly likely to be contiguous with the ventricles. In the most extreme case, with almost all white matter involved, up to 100% of WML would thus be labelled “periventricular”, although extending almost to the cortex; it is clear that under such conditions the term periventricular becomes meaningless and might be better abandoned. Apart from questionable anatomical definitions, it is clear that this bias will preclude finding any clinico‐radiological associations in the subcortical zone.

Since there is no radiologically meaningful definition on which part of the white matter is periventricular, and what constitute deep white matter, nor evidence for a distinct class of subjects with either type of lesion in isolation,⁷ it would make more sense to abandon this artificial classification and analyse WML in a more anatomically meaningful way. Using advances in computational neuroanatomy, this is now becoming feasible. One can either map a template (an atlas with known anatomical relationships) to the subject space to calculate a fractional volume of WML in a given tract, or study the relationship between a given cognitive measure and the probability of finding a WML across a group of subjects registered to standard space. On the other hand, the apparent controversy detailed above marks the need for better radiological‐pathological correlation studies, applying more sophisticated MRI and histopathological techniques.