Quantifying hubness to predict surgical outcomes in epilepsy: Assessing resection‐hub alignment in interictal intracranial EEG networks DOI
Ruxue Gong, Rebecca Roth,

Kaitlyn Hull

и другие.

Epilepsia, Год журнала: 2024, Номер 65(11), С. 3362 - 3375

Опубликована: Сен. 21, 2024

Abstract Objective Intracranial EEG can identify epilepsy‐related networks in patients with focal epilepsy; however, the association between network organization and post‐surgical seizure outcomes remains unclear. Hubness serves as a critical metric to assess by identifying brain regions that are highly influential other regions. In this study, we tested hypothesis favorable post‐operative associated surgical removal of interictal hubs, measured novel “Resection‐Hub Alignment Degree (RHAD).” Methods We analyzed Phase II intracranial from 69 epilepsy who were seizure‐free ( n = 45) non–seizure‐free 24) 1 year post‐operatively. Connectivity matrices constructed recordings using imaginary coherence various frequency bands, centrality metrics applied hubs. The RHAD quantified congruence hubs resected/ablated areas. used logistic regression model, incorporating clinical factors, evaluated alignment regarding outcomes. Results There was significant difference fast gamma (80–200 Hz) unfavorable p .025). This finding remained similar across definitions (i.e., channel‐based or region‐based network) measurements (Eigenvector, Closeness, PageRank). surgically removed areas commonly quantitative measures (seizure‐onset zone, irritative high‐frequency oscillations zone) did not reveal differences suggests hubness measurement may offer better predictive performance finer‐grained analysis. addition, showed explanatory validity both alone (area under curve [AUC] .66) combination therapy type (resection vs ablation, AUC .71). Significance Our findings underscore role hub removal, through high networks, enhancing our understanding surgery.

Язык: Английский

Deep learning on brief interictal intracranial recordings can accurately characterize seizure onset zones DOI
Sameer Sundrani, Graham W. Johnson, Derek J. Doss

и другие.

Epilepsia, Год журнала: 2025, Номер unknown

Опубликована: Май 27, 2025

Abstract Objective Epilepsy is a debilitating disorder affecting more than 50 million people worldwide, and one third of patients continue to have seizures despite maximal medical management. If patients' localize discrete brain region, termed seizure onset zone, resection may be curative. Localization often confirmed with stereotactic electroencephalography; however, this require stay in the hospital for weeks capture spontaneous seizures. Automated localization zones could therefore improve presurgical evaluation decrease morbidity. Methods Using 1 000 interictal electroencephalography segments collected from 78 patients, we performed five‐fold cross‐validation testing on multichannel, multiscale, one‐dimensional convolutional neural network classify zones. Results Across held‐out test sets, our models achieved zone classification sensitivity .702 (95% confidence interval [CI] = .549–.805), specificity .741 CI .652–.835), accuracy .738 .687–.795), which was significantly better trained random labels. The well across entire brain, top five region performance demonstrating accuracies between 70.0% 88.4%. When split by outcomes, favorable Engel outcomes after or who were responsive neurostimulation responders. Finally, SHAP (Shapley Additive Explanation) value analysis median‐normalized input data assigned consistently high feature importance spikes large deflections, whereas similar analyses histogram‐equalized revealed differences assignments low‐amplitude segments. Significance This work serves as evidence that deep learning brief intracranial can brain. Furthermore, findings corroborate current understandings epileptiform discharges help uncover novel morphologies. Clinical application reduce dependence recorded shorten time drug‐resistant epilepsy reducing patient morbidity costs.

Язык: Английский

Процитировано

0
Quantifying hubness to predict surgical outcomes in epilepsy: Assessing resection‐hub alignment in interictal intracranial EEG networks DOI
Ruxue Gong, Rebecca Roth,

Kaitlyn Hull

и другие.

Epilepsia, Год журнала: 2024, Номер 65(11), С. 3362 - 3375

Опубликована: Сен. 21, 2024

Abstract Objective Intracranial EEG can identify epilepsy‐related networks in patients with focal epilepsy; however, the association between network organization and post‐surgical seizure outcomes remains unclear. Hubness serves as a critical metric to assess by identifying brain regions that are highly influential other regions. In this study, we tested hypothesis favorable post‐operative associated surgical removal of interictal hubs, measured novel “Resection‐Hub Alignment Degree (RHAD).” Methods We analyzed Phase II intracranial from 69 epilepsy who were seizure‐free ( n = 45) non–seizure‐free 24) 1 year post‐operatively. Connectivity matrices constructed recordings using imaginary coherence various frequency bands, centrality metrics applied hubs. The RHAD quantified congruence hubs resected/ablated areas. used logistic regression model, incorporating clinical factors, evaluated alignment regarding outcomes. Results There was significant difference fast gamma (80–200 Hz) unfavorable p .025). This finding remained similar across definitions (i.e., channel‐based or region‐based network) measurements (Eigenvector, Closeness, PageRank). surgically removed areas commonly quantitative measures (seizure‐onset zone, irritative high‐frequency oscillations zone) did not reveal differences suggests hubness measurement may offer better predictive performance finer‐grained analysis. addition, showed explanatory validity both alone (area under curve [AUC] .66) combination therapy type (resection vs ablation, AUC .71). Significance Our findings underscore role hub removal, through high networks, enhancing our understanding surgery.

Язык: Английский

Процитировано

1