Abstracts

Rationale:
Functional neuroimaging enables exploration of brain networks that support cognitive functions and features that are associated with good cognitive outcome following resective surgery for seizure control. Developing clinical fMRI tools to predict post-operative changes in memory following anterior temporal lobe resection (ATLR) remains an important goal of this enterprise. Here, we evaluated the utility of preoperative resting state fMRI (rs-fMRI), comparing patients to a ‘normative’ sample, to predict post-operative verbal memory decline following left ATLR.
Method:
Twenty nine patients with left temporal lobe epilepsy (TLE) and 19 healthy controls underwent 6 minutes of resting state functional and structural MRI. Patients underwent neuropsychological testing before and after ATLR for treatment of intractable seizures, and a composite score of verbal memory capacity was derived. A set of regions-of-interest were derived from meta-analyses (Brainnetome Atlas) to specify a canonical ‘memory network’. In order to assess the normalcy of each patients’ memory network (which may reorganize with epileptic activity), we compared each patient’s preoperative resting-state memory network connectivity to ‘normative’ healthy control network connectivity, and used the resulting measure of network similarity to predict postoperative verbal memory change. In parallel, we analyzed graph theory degree within to-be-resected regions to examine if the extent of integration of the anterior temporal lobe with the broader memory network preoperatively was predictive of verbal memory decline after resection, providing a more focal measure of the pathological temporal lobe contribution to memory. Thus, we analyzed network-wide and focal properties associated with cognitive morbidity and resilience.
Results:
We found that patients with connectivity patterns that were more similar to the normative template showed less verbal memory decline (r=0.35, p=0.03), indicating that ‘weaker’ or reorganized networks are more functionally vulnerable to network insult with temporal lobe resection. In tandem, we found that individuals with to-be-resected regions that were more integrated with the memory network pre-surgery showed greater memory decline after resection (r=0.43, p=0.009), indicating that removing more functionally active hubs within the memory network is detrimental to outcome. 
Conclusion:
These findings indicate that six minutes of preoperative resting state scanning can inform postoperative verbal memory decline in patients with left TLE, characterized by a single scalar value (correlation of overall network to the normative template).  This is important as memory activation tasks may be inappropriate or compromised in some circumstances or patients.  Our findings also underscore the conclusion that both focal measures of temporal lobe functioning as well as broader network organization contribute to memory capacity in the context of temporal lobe epilepsy.
Funding:
:
Funding:
for this study was provided by the Ontario Brain Institute (EpLink program) and the Canadian Institutes of Health Research.