Thalamic amnesia after infarct: The role of the mammillothalamic tract and mediodorsal nucleus

Danet et al.1 are commendable for quantitatively associating thalamic lesions with memory deficits. Previous evidence reported a significant association between mediodorsal parvocellular lesions and memory deficits not replicated in this article.2 However, some procedures require more details. The clinical group showed significant nonmnemonic impairments (table 2),1 which entangles comparisons with previous studies.2,3 It is unclear how psychiatric history of participants, drug abuse, and pharmacologic treatment were considered. Techniques accounting for postlesion shrinkage accurately reproduce radiologic inspection,4 but I am not aware of a validation of the technique used by Danet et al. Validation of technique is critical to clarify how the lacunae were weighed and warped in the normalization procedure. The atlas masks employed overlap at the borders,5 with voxels belonging to multiple masks; it is unclear how this confound was overcome. It is also unclear how many lesion–behavior correlations have been performed and corrected for multiple comparisons. Clinical subgroups differ by mammillothalamic tract (MTT) volume, overall lesion size (Mann-Whitney U test, p 5 0.018), and ventral nuclei volumes (p 5 0.03). Surprisingly, MTT volumes correlate just moderately with z-transformed scores (table 3; Spearman r 5 20.52, p 5 0.08). These points should be considered when interpreting the results reported.