Abstracts

Using a set of morphological characteristics that were visually assessed on MRI, we previously showed (Bernasconi, Brain, 2005) that abnormalities of shape and positioning of the hippocampal formation (HF) were more frequently found in TLE patients than healthy controls. Here, our purpose was to model quantitatively shape and positioning characteristics of the hippocampal formation using mathematical descriptors., We studied 44 healthy controls and 30 patients with TLE (17 left TLE and 13 right TLE), including 15 with hippocampal atrophy (HA) and 15 with normal hippocampal volumes (NV), as determined by MR volumetry. Mathematical descriptors were calculated on the manually segmented hippocampi. The following 2D descriptors were measured at the level of the hippocampal head, body and tail: medial positioning (MP), roundness (RO), and vertical orientation (VO). Medial positioning and vertical orientation were also measured by calculating axial and longitudinal rotation using 3D models estimating angle differences between the principal axes of a target hippocampus and those of a reference average of 152 controls. HF shape and positioning were considered abnormal if at least 2/3 2D descriptors and 1/2 3D descriptors were +/- 2SD from the mean of controls., In controls, compared to the visual assessment (3/44=7%), both 2D and 3D models allowed detection of more HF abnormalities (7/44=16% and 8/44=18%, respectively). In TLE, results for 2D models were similar to the visual analysis (27% vs. 33%). However, 3D models showed higher percentage of abnormalities (57% vs. 33%; p[lt]0.001). Models identified all but one subjects that were considered abnormal in the visual assessment. Group analysis for 2D models showed that in controls the left hippocampus was rounder and more vertically oriented than right hippocampus (p[lt]0.01). Compared to controls, in TLE, the examined 2D shape and positioning abnormalities were also more common in the left hippocampus (p[lt]0.05) and occurred in both patients with HA and NV. Moreover, there was a significant asymmetry (p[lt]0.05) and variance ([italic]F[/italic]-test, p[lt]0.001) in 3D axial and longitudinal rotation (L[gt]R). Abnormal features were not related to the side of the focus., Models of HF shape and positioning are reliable and more sensitive than visual evaluation. In both controls and TLE patients, HF abnormalities occurred more frequently than on visual assessment. These abnormalities predominate in the left HF and are not related to the side of the focus in TLE.[figure1], (Supported by the Canadian Institutes for Health Research.)