dimensions. In all cases the value of F3 appears to reduce with time, indicating a brain that tends to become less organized, with its indicatrix becoming smoother and rounder. The only exception appeared to be that of case (g) coming from a cavernous hemangioma sufferer. However, this case does not correspond to a diffuse pathology; instead it contains a localised tumor. When the region of the tumor alone is considered, the value of F3 again reduces from the first scan to the second.
FIGURE 12 Features computed for seven MR-t2 images: 1,2 from normal subjects; 3,4 from Alzheimer's sufferers; 5 from a vascular dementia sufferer; 6 from a chronic subdural hematoma sufferer; 7 from an AIDS dementia sufferer. In the left column we plot Fl and in the right F2. (a) Features computed from the GD orientation histograms for the whole brain component. (b) Features computed from the INV indicatrices for the whole brain component. (c) Features computed from the GD orientation histograms for the interior component of the brain only. (d) Features computed from the INV indicatrices for the interior component of the brain only.
texture of the brain when pathology is present. This is observed when either of the two features is used (F1 presented on the left, F2 on the right) and when the analysis is performed either for the whole brain (Fig. 12a) or for the darkest component only (Fig. 12c). The results obtained with the INV method do not show as much consistency for this particular data set.
Finally, in Fig. 13 we present some results of calculating feature F3 from the GD orientation histograms of 14 images of various modalities, referring to seven different patients with various conditions and forming pairs obtained a few months apart. The figure shows some typical slices of these images, although the actual analysis has of course taken place in three
Was this article helpful?