Quantitative Analysis of Longitudinal Memory fMRI in Temporal Lobe Epilepsy Using ICN_Atlas

1345 

Abstract Submission 

Azka Sohail¹, Marine Fleury¹, Meneka Sidhu¹, Lajos Kozák², Louis Lemieux¹

¹UCL Queen Square Institute of Neurology, London, United Kingdom, ²Department of Neuroradiology, Medical Imaging Centre, Semmelweis University, Budapest, Hungary

Azka Sohail  
UCL Queen Square Institute of Neurology
London, United Kingdom

Marine Fleury  
UCL Queen Square Institute of Neurology
London, United Kingdom

Meneka Sidhu, MD, Ph.D  
UCL Queen Square Institute of Neurology
London, United Kingdom

Lajos Kozák, MD, Ph.D  
Department of Neuroradiology, Medical Imaging Centre, Semmelweis University
Budapest, Hungary

Louis Lemieux, Ph.D  
UCL Queen Square Institute of Neurology
London, United Kingdom

The impact of surgery on memory function in severe epilepsy patients and its prediction have been investigated using functional magnetic resonance imaging (fMRI). Notably, strong associations between fMRI lateralisation and postoperative verbal memory decline were observed (Sidhu et al., 2015). In a related study, Sidhu et al. (2016) assessed surgical success by examining pre- and postoperative memory reorganisation following anterior temporal resection. Traditionally, these investigations frame findings based on anatomical localisation, describing activated regions and selecting seed regions for functional connectivity analyses. Recent research has highlighted how Intrinsic Connectivity Networks (ICNs) capture fundamental aspects of the brain's functional connectivity, offering a universal framework for fMRI interpretation and quantification of activation patterns as a basis of fMRI map regional quantification. We have implemented this approach in the form of the ICN_Atlas (Kozák et al. (2017)) to quantify the degree of engagement of ICNs in fMRI maps (figure 1).

In this study, we aimed to quantify changes in memory encoding activation patterns pre and post-surgery using ICN_Atlas, introducing an objective approach to supplement previous investigations. 41 patients with temporal lobe epilepsy (20 left TLE, 21 right TLE) were scanned pre-operatively and post-operatively at 4 months. 20 healthy controls underwent the same fMRI protocol. ICN engagement was quantified using two metrics: Spatial Involvement (IRi) and Normalised Mean Activation (MAN,i) for 20 ICNs (Laird et al., 2011). We assessed the ICN-wise engagement changes using repeated measures ANOVA.

A representative pattern of change is illustrated in Figure 2. In the healthy controls, the following patterns were observed for both tasks: no significant change in IRi and with an initial modest increase in MAN,i followed by a decline at 10 years. In the patient groups, the most prominent changes following surgery were observed in the motor/visual-spatial and visual ICNs. Notably, at 3 months' post-surgery, motor ICNs exhibited increased spatial engagement while visual ICNs showed little to no change in both patient sub-groups. At 10 years post-surgery, spatial engagement of the motor ICNs declined markedly, contrasting with a substantial increase in visual ICNs. The patterns for MAN,i show a pronounced change in ICN engagement, particularly for word encoding, in contrast to IRi. This indicates there are substantial changes in activation strength across different rescans. Overall, the engagement changes were more significant in the patients than in the healthy controls and greater in LTLE compared to RTLE.

The observed variations in ICN engagement are in line with expectations for memory encoding tasks, signifying the reorganisation of memory encoding networks. Utilising ICN_Atlas, we were able to quantify longitudinal fMRI data objectively in a whole-brain, regionally specific and functionally meaningful manner. The quantitative perspective offered here complements the anatomical focus of previous investigations (Sidhu et al., 2013, 2015, 2016) and, therefore, could lead to a better, more complete understanding of the impact of surgery and, potentially, improved prediction of functional sequela.

This work is supported by the UCLH BRC and MRC grant ID MR/X031039/1 who also support MKS.

Higher Cognitive Functions Other

Long-Term Memory (Episodic and Semantic)

Activation (eg. BOLD task-fMRI) ¹

Segmentation and Parcellation ²

Brain Atlases

Atlasing

Cognition

Computational Neuroscience

Data analysis

Design and Analysis

Epilepsy

FUNCTIONAL MRI

Memory

Open-Source Software

Segmentation