Postdoctoral Fellow The University of Alabama at Birmingham Birmingham, Alabama
Rationale: Hippocampal dentation (HD) is a feature of hippocampal morphology that refers to surface convolutions seen in the CA1/subiculum region of the inferior aspect of the hippocampus. HD has been found to vary in prominence from quite bumpy to quite smooth in healthy controls (Figure 1), and is positively correlated with memory performance . However, up to now HD has been quantified only using subjective visual rating systems. Further, manual HD quantification is time consuming and prone to inconsistency across raters. Therefore, an objective measure to quantify HD is needed to advance study of this morphologic feature. Methods: Fifty-nine cases with hippocampal sclerosis (HS) from a retrospective database were included. Using a novel segmentation approach, standard T1w images with a resolution of 1-1.2mm3 were resampled to 0.25mm isotropic, and segmentation of the hippocampi was performed using Automatic Segmentation of Hippocampal Subfields . Surface meshes were generated from the segmentations followed by Taubin smoothing. Using MeshLab, a “swath” containing the entirety of HD was selected from each whole hippocampal surface mesh (Figure 1), and reoriented to align its principal axis in the x-y plane. A strip was traced on the swath to get a set of points that reflects the essential topology of HD, and further simplified to get a single curvilinear set of points. Low pass filtering was performed to extract the general shape of the hippocampus and was subtracted from the simplified strip to create a curve that reflects only undulations in surface contour (i.e., HD). Area under the curve (AUC) was calculated from the difference data to quantify HD (Figure 2). Intra-/inter-rater reliability measures were calculated using intraclass correlation coefficients (ICC). Student’s t-Test was performed to detect a difference between the epileptic and non-epileptic hippocampi. A Pearson’s Correlation Coefficient was calculated to test the correlation of AUC with visual count of dentes and neuropsychometric testing (NPT) measures including the Rey Complex Figure Recognition Test, California Verbal Learning Test (CVLT)-II trials 1-5 and CVLT-II long delay free recall. Results: Our novel segmentation approach demonstrates HD much more clearly across the subjects (Figure 1) than native resolution segmentations. The intra- and inter-rater ICC was ≥0.99 for three repeated AUC measurements across two raters. The AUC values were found to be significantly different (p=0.00019) between epileptic and contralateral hippocampi and significantly correlated with visual count of dentes (p< 0.05). Among the left HS cases, CVLT-II trials 1-5, which reflects efficiency of verbal memory acquisition, showed a positive trend with AUC values of right HD (p=0.07), but was not statistically significant. Other analyses were not significant. Conclusions:
This is the first objective measure of HD.
We demonstrate that HD is asymmetrically diminished in HS, implying HD is a novel imaging hallmark of HS.
A non-significant trend suggests that prominent HD on the right is associated with preserved memory function in left HS, supporting the “functional reserve” hypothesis and possible reorganization of verbal memory function.