Mapping surface meshes into new subjects' scans via fluid registration. (a) N image volumes (subsequently called atlases) are randomly selected from the sample and the lateral ventricles are manually traced and converted into surface mesh models. (b) N new ventricular models are then produced by fluid registration of each image volume to a different atlas. (c) Surface points are converted into 3D parametric surface meshes composed of spatially uniform triangular tiles. (d) The N surface meshes are integrated by simple mesh averaging for each individual subject. (e) Medial curves (red) are extracted, and the radial distance of each ventricular boundary point to a medial curve may be interpreted as a local thickness. (f) These distance measures are then averaged across subjects at each boundary point and plotted in color to produce a regional measure of radial expansion or contraction of the ventricles.

Significance maps for correlations between local ventricular enlargement and (1) diagnosis (MCI vs. normal, AD vs. normal and AD vs. MCI); (2) cognitive scores (MMSE, Global CDR, and Sum of Boxes CDR); (3) Geriatric Depression scores; (4) Logical Memory Delayed Recall scores and (5) Hachinski Ischemic scores, adjusting for age, sex and educational level. The bottom panel shows the FDR-corrected p-values for each map; these are all highly significant, partly due to the very large sample sizes. Depression scores are linked with ventricular differences in several regions affected in MCI.

Partial correlation coefficients (r-maps) for the 3 diagnostic comparisons, showing the strength of association between radial ventricular size and diagnosis, as well as with cognitive and clinical scores. The correlations in the MMSE and Delayed Logical Memory map are negative (red colors) because a higher MMSE and Delayed Logical Memory score is associated with less degeneration (opposite to all the other ones). These maps are visually in very strong agreement with the corresponding p-maps, and so they are not shown for the other covariates.

(a) Correlations between local ventricular enlargement and CSF biomarker levels, including Aβ_1-42, pTau_181p, Tau, and ratios of Tau/Aβ_1-42 and pTau/Aβ_1-42 in the pooled data and within the AD group. The bottom panel shows the FDR corrected p-values (critical p-values) for these maps. The critical p-value is the highest threshold that can be applied to the statistical map while keeping the false discovery rate below 5%. Correlations detected on one side only probably reflect limited power to detect a small effect size on both sides; it is unlikely that these maps show any true laterality (other than the fact that the left occipital horn is about 5 mm longer, on average), as ventricular expansion on both sides is highly correlated. n.s. means not significant. (b) Significance maps show regions where differences in baseline ventricular anatomy are associated with subsequent decline, over the following year, in 3 commonly used clinical scores with the control for age, sex and educational level. As hypothesized, ventricular expansion is a better index of future changes in global cognition function than changes in more specific cognitive domains, which likely depend on the integrity of specific cortical and hippocampal regions (e.g., the cingulate for apathy, etc.). The bottom panel shows the FDR-corrected p-values for these maps.

(a)Significance maps for correlations between local ventricular enlargement and ApoE genotype. The bottom panel shows FDR corrected p-values (critical p-values) for these maps; n.s. means not significant. ApoE4 carriers versus non-carriers showed weak but significant differences. (b)Significance maps show the strength of association between radial ventricular size and ADAS-Cog tests, including word recognition, spoken language and word finding. The bottom panel shows the FDR corrected p-values for these maps. Associations were detected for all three ADAS-Cog tests.

Cumulative Distribution Functions (CDFs) for significance maps associating ventricular enlargement with clinical measures and CSF biomarkers. Red values (right) show the percentage of the maps with significant correlations, when a threshold is applied to control the false discovery rate at 5%. This threshold (the highest one that controls the FDR at 5%) is sometimes called the critical P-value. At that critical P-value, the volume of suprathreshold statistics is more than 20 times that expected under null-hypothesis, and can as high as 100% if every surface location shows correlations with the clinical or pathological measure (Chou, et al., 2009b, Hua, et al., 2009, Hua, et al., 2008a, Morra, et al., 2009). A-Beta shows strong correlations with ventricular size (72% of the surface shows correlations), while correlations with other CSF biomarkers were not detectable. By contrast, all cognitive measures correlated very highly with ventricular expansion, with more specific clinical ratings (such as depression and Hachinski Ischemic scores) showing slightly less effect sizes than more general clinical ratings (such as MMSE and CDR; 98-99% of the ventricular surface was correlated with these measures). Also, ApoE4 genotype and the ADAS-Cog tests including word recognition, spoken language and word finding all correlated with ventricular expansion.