University of Twente Student Theses
Structural and effective brain networks in focal epilepsy
Jelsma, Susanne (2022) Structural and effective brain networks in focal epilepsy.
PDF
16MB |
Abstract: | Introduction: Epilepsy is nowadays regarded as a network disorder instead of a focal disease. While epilepsy surgery is currently based on the removal of a local focus, a network approach seems more suitable and might eventually improve the surgical outcome. Insight into how epilepsy alters the patient-specific brain network is necessary to establish a network based surgical strategy. There are different approaches to characterize brain networks but it is unclear if and to what extent these approaches relate. Effective networks are described by the causal influence between brain areas by invasive perturbation of one of the areas with for example single pulse electrical stimulation (SPES). Structural networks are described by the anatomical connections between brain areas via white matter tracts and can be determined non-invasively with diffusion weighted imaging (DWI). Exploring the relation between structural and effective networks could deepen our understanding of epileptogenic networks by revealing the biologically plausible structural pathways that give rise to effective connections. DWI based structural network characterization is non-invasive and could be adapted earlier in the surgical trajectory than SPES based effective network characterization. The combination of structural and effective networks could potentially elucidate network alterations caused by epilepsy. We aimed to characterize structural networks with DWI and effective networks with SPES, evaluate their relation, and explore how epilepsy alters this relation. Methods: We compared effective networks acquired by SPES to structural networks derived from DWI. Invasive electroencephalography (iEEG) electrode positions were used as nodes. Early responses (ER) in SPES were automatically detected, for which an existing detection algorithm for ERs in electrocorticography (ECoG) was optimized and validated for stereo EEG (sEEG) (Chapter 2). The optimized detector was used to reconstruct the effective networks. We reconstructed structural networks from DWI and fiber tractography(FT) using the iFOD2 algorithm with parameters optimized for local network structures (Chapter 3). We determined the inter-modal similarity between structural and effective networks with the Jaccard index (JI). We compared the topography with the degree and betweenness centrality on electrode contact level within patients. We constructed a linear multilevel model to evaluate the correlation at group level, accommodate for node proximity bias due to irregular spatial sampling, and analyze the influence of epilepsy. Results: We included 13 patients (five ECoG, eight sEEG). The sensitivity and specificity of the optimized ER detector were 81% and 93%. The FT algorithm for sEEG and ECoG required equal parameters. The median JI was 0.25 (IQR: 0.19-0.29). The degree of the structural networks compared to the effective networks at patient level showed a significant positive correlation in 10/13 patients. This correlation was also present at group level with linear multilevel modeling after controlling for node proximity. We did not find statistical evidence that epilepsy alters the relation between structural and effective networks. Conclusion: We explored the relation between structural and effective patient-specific brain networks. The performance of the optimized automatic ER detector was sufficient to reliably characterize effective networks. Structural and effective networks showed a moderate overall relation and their topography, described by the degree correlated independently of common sources of bias. This suggests that for some applications structural and effective networks could be interchangeably used, in which case we recommend to use non-invasive, structural networks. Higher sample sizes and correction of the node proximity on a patient level are needed to exactly explain to what extent structural and effective networks interrelate and further investigate how epilepsy alters this relation. We recommend continuing a multi-modality approach to study complex network alterations in focal drug resistant epilepsy patients to establish a network based surgical strategy. |
Item Type: | Essay (Master) |
Clients: | UMC Utrecht |
Faculty: | TNW: Science and Technology |
Subject: | 30 exact sciences in general, 44 medicine, 50 technical science in general |
Programme: | Technical Medicine MSc (60033) |
Link to this item: | https://purl.utwente.nl/essays/93515 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page