Abstracts

Rationale: Hippocampal internal architecture (HIA) asymmetry has been observed in patients with temporal lobe epilepsy (TLE) and has been demonstrated to have utility in predicting laterality of seizure onset when other signs of hippocampal sclerosis are absent. Published methods of assessing HIA clarity involve visual scoring of HIA according to a four-point scale. While time-efficient and inexpensive, these rating systems are subjective. We developed a quantitative technique that assesses the same features and tested its performance. Methods: T2 weighted, high-resolution coronal slices through the hippocampal body were analyzed in 35 patients with ictal evidence of unilateral TLE (right=18, left=17) utilizing both a) the hippocampal internal architecture (HIA) rating scale (Ver Hoef et al 2013) and b) a new quantitative technique designed to measure the same information. The proposed technique identifies pixels in the internal portion of the hippocampus as representing the white matter band marking the border of the dentate gyrus and the hippocampal pyramidal cell layer if they have an intensity 2.5 standard deviations below that of the CA4 region in the same image or lower. A mean intensity is calculated for the hypointense pixels, subtracted from the mean pixel intensity of CA4 and divided by an estimate of the local image noise to calculate a contrast to noise ratio (CNR). The CNR for each slice is multiplied by the number of hypointense pixels to calculate a HIA conspicuity index, which is then averaged across slices on each side and subtracted left from right to generate conspicuity asymmetry index. A t-test was performed to determine if there was a significant difference in asymmetry scores between right and left TLE patients and a logistic regression model was utilized to determine if more extreme asymmetry scores were associated with a greater likelihood of correct lateralization. Results: A one-tailed Student's t-test showed a significant difference between the mean qualitative HIA rating scale scores in right and left TLE patients (p<0.00001) however only a non-significant trend was observed using the quantitative technique (p=0.187). Logistic regression analysis did not show the HIA conspicuity asymmetry to be a significant predictor of laterality of seizure onset (b=0.558, SE=0.607, p = 0.358), however, as previously shown, the qualitative HIA asymmetry score was a strong predictor of the laterality of seizure onset (b=3.7165, SE=1.3680, p=0.0066). Conclusions: While the proposed quantitative technique shows a trend toward demonstrating a difference between right and left TLE patients, it neither reaches statistical significance nor has the high accuracy and specificity of our previous qualitative method. The accuracy of this technique requires very high quality imaging data whereas the human reader employing the qualitative approach may perceive subtle patterns even in noisier images. Improvements in imaging quality are needed for unbiased, quantitative measures of HIA to be accurate.