Combined Functional and Structural Network Analysis for Predicting Language Decline in Anterior Temporal Lobe Resection
Abstract number :
2.277
Submission category :
9. Surgery / 9A. Adult
Year :
2022
Submission ID :
2204912
Source :
www.aesnet.org
Presentation date :
12/4/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:27 AM
Authors :
Lawrence Binding, BSc, MSc – UCL; Peter Taylor, PhD – CNNP lab, Interdisciplinary Computing and Complex BioSystems Group, School of Computing Science – Newcastle University; Zara Fenlon, BSc – Assistant Psychologist, Department of Neuropsychology, University College London Hospital; Lucy Roberts-West, BS, MSc – Assistant Psychologist, Department of Neuropsychology, University College London Hospital; Marine Fleury, BSc, MSc – Department of Clinical and Experimental Epilepsy – UCL Queen Square Institute of Neurology; Davide Giampiccolo, MD – Victor Horsley Department of Neurosurgery – National Hospital for Neurology and Neurosurgery; Lorenzo Caciagli, PhD – Department of Bioengineering – University of Pennsylvania,; Jane de Tisi, BA – Department of Clinical and Experimental Epilepsy, – UCL Queen Square Institute of Neurology; Andrew McEvoy, FRCS (SN) – Victor Horsley Department of Neurosurgery – National Hospital for Neurology and Neurosurgery; Anna Miserocchi, FRCS (SN) – Victor Horsley Department of Neurosurgery – National Hospital for Neurology and Neurosurgery; John Duncan, FRCP, FMedSci – Department of Clinical and Experimental Epilepsy – UCL Queen Square Institute of Neurology; Sjoerd Vos, PhD – Centre for Microscopy, Characterisation and Analysis – University of Western Australia
Rationale: Language decline occurs after 30-50% of anterior temporal lobe resections (ATLR). Damage to white matter fibre bundles can explain language decline and individual connections can predict picture naming decline with around 80% accuracy. Here, we advance prior work by creating a diffusion MRI and functional MRI (fMRI) fused model of connectivity, and aim to produce individualised patient models that can predict postoperative language decline._x000D_
Methods: We examined data from 15 TLE patients who had left-lateralized language and left-sided resections. Structural connectivity was derived from preoperative diffusion MRI using probabilistic tractography. Functional connectivity was derived via preoperative verbal fluency fMRI using the conn toolbox.[1] Structural postoperative connectomes were estimated from manually drawn resection masks based on pre-/post-operative 3D-T1 scans. Pre- to post-operative change in the structural network was calculated by dividing the post-operative by the pre-operative networks. Post-operative functional networks were estimated by multiplying the pre-operative functional network by the pre- to post-operative change in structural network. The pre- and post-operative structural network was then weighted with the pre- and post-operative functional network. Finally, the weighted post-operative network was divided by the weighted pre-operative network. We investigated three metrics of language ability: McKenna Graded Naming Test (GNT), semantic and phonemic fluency before and 3 months post-operatively. Pre- to post-operative language change was binarized using a reliable change index. A generalized linear model with SCAD penalization was used to identify features related to language outcome. Each reduced feature set was entered into a support vector machine incorporating a leave-one-out cross-validation scheme. The receiving operator characteristic area under the curve (AUC) metric was used to identify the best predictive model; average accuracy (ACC) was also calculated. _x000D_
Results: Two connections were associated with naming decline 1) MTG to pallidum 2) MTG to dorsal cingulate cortex (AUC=0.93, ACC=93%). Nine connections were associated with semantic decline, containing both intra- and inter-hemispheric connections, to the temporal, insula, and cerebellum cortex (AUC=0.66, ACC=73%). Two connections were associated with phonemic decline 1) Calcarine to anterior collateral sulcus 2) anterior to posterior collateral sulcus (AUC=0.90, ACC=87%)._x000D_
Conclusions: We demonstrate that the fusion of functional and structural imaging data is powerful in predicting postoperative language decline. These results will contribute to creating patient-bespoke surgical plans, avoiding tracts and/or areas that are crucial for function in individual patients.
Funding: Epilepsy Research UK (grant number P1904), Wellcome Trust Innovation Program (218380/Z/19/Z), National Institute for Health Research University College London Hospitals Biomedical Research Centre (NIHR BRC UCLH/UCL High Impact Initiative BW.mn.BRC10269)
Surgery