Abstracts

Rationale: Memory impairment is a risk following epilepsy surgery. Without a reliable memory fMRI task, pre-operative language fMRI may help to characterize mesial temporal lobe (MTL) functioning for children (Sepeta et al. 2016) and potentially aid in predicting post-operative memory functioning. The utility of language fMRI for this purpose has been shown in adults (Binder et al. 2008, 2010), but has not been examined in pediatric epilepsy. In this study, we aim to investigate if the overlap between resection and activation during language fMRI predicts post-operative memory decline in pediatric patients.

Methods: 17 epilepsy patients (ages 5.5-15.6; Table 1) completed neuropsychological evaluation of memory (Table 2) before and after surgery. Difference scores (post-surgical minus pre-surgical) were calculated to examine post-surgical change. Patients completed pre-operative language fMRI (auditory decision description task) to determine MTL and language laterality using LI Toolbox. We automatically generated resection masks and computed the overlap amount of activation (magnitude) within the memory network (Neurosynth “episodic memory” mask, including hippocampus, parahippocampus, posterior cingulate, and additional limbic, frontal, and parietal regions). We used linear regression to predict post-operative memory decline from this overlap and the following factors: MTL LI, language dominance, and side of resection in the full sample (n=17) and temporal resection group (n=11).

Results: Pre- and post-operative memory scores were not different (all p’s > 0.05); however, 8/17 declined on at least one measure. All 17 patients had significant activation within the memory network during the auditory description decision fMRI task. 13 patients (76%) had overlap between surgical resection and memory network activation. The amount of activation resected did not predict memory change. Examining only temporal resections, 10 patients (90%) demonstrated overlap between resection and memory network activation. This overlap predicted post-operative decline on CVLT long delay (R2=0.54, p=0.01) stories immediate (R2=0.54, p=0.01), and CVLT short delay at trend level (R2=0.35, p=0.055). MTL LI, language dominance, and side of resection did not predict memory outcome for either group (full sample or temporal).

Conclusions: In the larger surgical sample, 76% of children had resected activation in the memory network during language fMRI. The amount of this activation predicted memory outcome for temporal resections only. No other variables—MTL LI, language dominance—aided in prediction, indicating amount of memory network activation resected during language fMRI may specifically predict post-operative memory in this set of patients. Language fMRI may provide an assessment of risk for memory impairment when resections within activated regions of the memory network are pursued for seizure freedom.

Funding: Please list any funding that was received in support of this abstract.: K23 NINDS NS093152; NINDS Division of Intramural Research.