Figure 1: Shape averaging reveals a significant shape difference in the corpus callosum of schizophrenic patients relative to controls, with clear differences in average anatomy between male and female schizophrenic patients. Maps are based on a high-resolution magnetic resonance image database (15 patients, 15 matched normal subjects; all male). [Data from Narr et al., 2000; for computational approach, see here]
Figure 2: Schizophrenic patients display a far greater anatomical variability (red colors) in the frontal cortex, than control subjects matched for age, gender, and other demographic factors. This indicates an aberrant organization of the gyral pattern in frontal cortex, perhaps occurring during late embryonic development, when the gyral pattern is established.
Figure 3: In schizophrenia, the pattern of greater anatomical variability is specific to frontal cortex, and is found in both male and female patients (SZ) but not in normal controls (NC). [Data from Narr et al., 2000; for computational approach, see here]
Figure 4:
Average anatomical shapes are shown for 3 subject groups:
10 schizophrenic patients, 10 normal subjects, and 10 patients with
clinically-determined Alzheimer's
Disease. In the schizophrenic group, the bowing of the corpus
callosum is most prominent anteriorly (right hand side of this
picture). Here the anterior callosum acts as the
superior boundary of the 3rd ventricle, and the entire ventricular
system is often enlarged in schizophrenic patients (see Figure 8,
below). In
dementia, the bowing of the corpus callosum occurs more posteriorly, and
is accompanied by a 25% focal fiber loss at the isthmus (left).
[Data from Thompson et al., 1998, and Narr et al., 1999].
Figure 5: The corpus callosum is automatically detected in an image by a computer vision algorithm which has learned the parameters of normal shape variability from a population of human subjects. A deformable curve is evolved to maximize a pre-defined measure of fit, which is based on (1) agreement with a diffused edge map, (2) regularity of the contour boundary, and (3) statistical deviation from a database of normal callosal shapes. [Data from Pitiot et al., 1999 and Thompson et al., 2000].
Figure 6: This map represents the magnitude and principal directions of neuroanatomical variability in normal populations (pink brain areas, highest variability, blue areas, little variability). These probability clouds are computed from shape transformations. The resulting probabilistic model of the human cortex can be used to determine whether anatomical variants in a patient are within the normal range. [Data from Thompson et al., 2000].
Figure 7: A computer vision algorithm compares cortical patterns between a normal subject and a patient by warping models of their brain surfaces and computing the resulting shape differences. [Data from Thompson et al., 2000].
Figure 8: The most prominent structural difference in schizophrenic brains is a substantial enlargement of the lateral ventricles, especially in posterior regions. These average anatomical models show a prominent enlargement in the occipital horns. The occipital horns are also the most anatomically variable area of the ventricular system (red colors, highest variability).
