Authors :
Presenting Author: Lawrence Binding, BSc, MSc – UCL
Davide Giampiccolo, MD – UCL; Yaqi Ji, BSc, MSc – University of California; Marine Fleury, BSc, MSc – UCL; Gavin Winston, BM BCh, PhD, FRCP – Queens University; Aidan O'Keeffe, PhD – University of Nottingham; Andrew McEvoy, FRCS (SN) – UCL; Anna Miserocchi, FRCS (SN) – UCL; Peter Taylor, PhD – Newcastle University; John Duncan, FRCP, FMedSci – UCL; Sjoerd Vos, PhD – The University of Western Australia
Rationale:
Temporal lobe resection for medically refractory temporal lobe epilepsy (TLE) can result in a substantial episodic memory deficit. Verbal episodic memory decline is often at greater risk following language-dominant resection. Episodic memory relies on the mesial temporal lobe-to-thalamic axis which is supported by direct and indirect subcortical pathways. The fornix subserves the direct pathway and interconnects sub-cortical structures with the hippocampus. The indirect pathway is less understood. However, research suggests that the parahippocampus plays a key role in verbal episodic memory decline following temporal lobe resection. This work examines the relationship between the transection of direct and indirect pathways to post-operative changes in verbal recognition and delayed recall episodic memory.
Methods:
Reconstruction of the ventral cingulum (vCing) and medial occipital longitudinal tract (MOLT) for the indirect and the fornix for the direct pathway was performed using pre-operative diffusion MRI. Resection masks were drawn comparing co-registered pre and post-operative 3D-T1 MRI scans. Percentage of resection was estimated by using resection masks as exclusion masks. One hundred and twenty-seven patients’ verbal recognition and delayed recall neuropsychology scores were z-scored against healthy controls pre-operatively and 3 and 12 months post-operatively. Patients were dichotomized into language-dominant and non-dominant resections via functional MRI. Disconnectome analysis and proportion of fiber bundle transection analysis were performed with linear mixed effect models to assess transection impact on verbal memory. Machine learning was used to evaluate the predictive capability of the pre-operative fiber bundle's microstructure and outcome.
Results:
Patients with language-dominant resections had significantly greater decline in verbal recognition and delayed recall scores at three and twelve months compared to non-dominant resections. Fibre bundles were only significantly related to outcome on the language-dominant hemisphere. Disconnection of the fornix, MOLT, and vCing was significantly related to a greater decline in z-scores at three and twelve months for verbal recognition and delayed recall. More extensive resection of the MOLT was significantly related to a greater decrease in verbal recognition z-scores at three and twelve months (Figure 1). Greater resection of the fornix was significantly related to greater decrease in verbal delayed recall scores at three and twelve months (Figure 2). The microstructure of these fiber bundles could predict a patient's z-score change with a unified mean absolute error of 0.88 for verbal recognition and 0.94 for verbal delayed recall.
Conclusions:
Our results suggest that verbal memory deficits may be caused by disconnection. While the fornix’s involvement is in line with early animal and human studies, this data indicates that both the ventral cingulum and the MOLT may contribute to memory circuits and their avoidance may improve memory outcomes in patients undergoing temporal lobe epilepsy surgery.
Funding:
This work is supported by Epilepsy Research UK (grant number P1904).