0.06 0.08 0.10 0.12 0.14 0.16 0.18 enlargement, region of isthmus
Figure 4.8. Enlargement of Figure 4.4 and Figure 4.5 (right) in the region of the isthmus. The reshaping of the lower arc here is responsible for most of the statistical significance of the overall group difference. A Procrustes fit of just these eight points results in an even clearer localization of the shape difference between these two groups.
on callosal shape. One is in genu: a generalized shrinking of the feature. (In the data set at hand, this effect should not be given too much credence, as the original slices from which these outlines were extracted were so thick that the region of rostrum was not reliably discernible.) More interesting is the change toward the posterior, in the splenium and the arc of isthmus just above. The underside of the isthmus appears to be substantially displaced in an upward-posterior direction. The difference is rather more one of location than one of area.
With this hint in hand, we know where to look. Drawing from the original scatter, Figure 4.8 enlarges this interesting region where the isthmus springs from the splenium. At this scale, it is possible to indicate the group origin of each outline point: The doctors' data points are shown by circles, the patients' by crosses. At the semi-landmark that has shifted most in Figure 4.8—second from the left in the lower segment—the groups are separated in the direction normal to the curve by a very substantial amount. Discarding the rest of the information in the outline, Bookstein (1997c) repeats the Procrustes analysis using only this subset of eight points and shows that in the resulting projection the doctors and the patients are separated at this locus with only two classification errors. The separation is by 1.75 times the standard deviation of the normative subsample, the largest effect I have ever seen reported in any neuroanatomical study of schizophrenia. The statistical significance of the overall comparison of mean outline shapes owes wholly to this sharp shape difference in one short arc of the ventricular boundary of the isthmus. Recall that was an overall significance test, a test for the presence of any substantial difference of location along any subarc or set of subarcs. The significance of any difference is tested first, globally. It is the overall level, not any Bonferroni correction of local findings, that applies to the study as a whole.
Was this article helpful?