Normalized Space and Time Improves Post-Operative Outcome Prediction in Epilepsy

Peter N. Taylor, PhD, staff researcher, Faculty of Medical Sciences and School of Computing Science, Newcastle University

Peter N. Taylor, PhD

The results of a retrospective study of 55 patients with refractory focal epilepsy who underwent epilepsy surgery suggest that normalizing for spatial proximity between nearby electrodes can improve the accuracy of post-surgery seizure outcomes.

Conducted by Peter N. Taylor, PhD, staff researcher, Faculty of Medical Sciences and School of Computing Science, Newcastle University, and colleagues, the study showed that patients with more extensive electrode coverage were more likely to have their outcome predicted correctly, but not necessarily more likely to have a better outcome.

“Future studies should account for the spatial proximity of electrodes in functional network construction to improve prediction of postsurgical seizure outcomes,” Taylor et al. wrote. “Greater coverage of both removed and spared tissue allows for predictions with higher accuracy.”

All told, the investigative group sought to address 3 important challenges identified in the use of interictal intracranial electroencephalography (iEEG) to predict postoperative seizure freedom, despite its relative success. Those included the spatial bias caused by the natural correlation of electrodes that are closer to each other; the difficulty of cross-subject comparisons due to the variety in electrode number and implant location; and assumptions of the stagnant nature of functional correlation networks despite their differences.

To do so, they used a measure termed D_RS, which they noted stands for the “distinguishability of the removed node strengths versus the spared node strengths” and has a single value per patient. As such, a D_RS value equal to 1(0) indicates all spared electrode contacts have a higher (lower) node strength than all removed electrode contacts.

WATCH NOW: FREEDOM Trial: Perampanel Monotherapy for Recurrent POS

When assessing the impact of spatial normalization to improve the discrimination between groups, they found that the original area under the receiver operating characteristic curve (AUC) revealed poor discrimination between outcome groups (AUC, 0.57; P = .037). After spatial normalization, patient groups show significant differences in their D_RS values, which discriminated outcome groups (AUC, 0.70; P = .02). They did note that the spatial normalization procedure could be improved going forward via accounting for and clearer understanding of biophysical and biological factors such as lobe‐ or region‐specific functions.

“Spatially under-sampling networks can directly lead to changes in the estimated network properties, and thus we investigated the impact of spatial sampling on our ability to distinguish outcome groups,” Taylor and colleagues wrote. In the data, they used 2 patients as an example: patient 985, who had 27 electrodes in both the removed and spared tissue, and patient 865, who had 9 electrodes in the spared tissue.

By increasing the coverage of the removed and spared tissue (n_x), the distinction between outcome groups was clearer. For n_20­, the analysis included 27 total patients, which resulted in an AUC of 0.91 (P = .0006).

“For practical applications using interictal iEEG functional networks to delineate epileptogenic tissue, it is important to understand whether and how our results change if different underlying data are used,” they wrote. To assess this, Taylor et al. explored if the main findings were impacted by duration and timepoint differences between functional networks. They found that typically, a 10‐second segment is not significantly worse than a 1‐hour segment over all n_x values, but the AUC varied more from segment to segment for segments of 10 seconds or less. This, they wrote, “[indicates] that consistency of results may drop for short segments.”

They noted that while the timescale and timepoint did not dramatically impact the results, that finding should not be interpreted as evidence of the stability of the interictal functional networks. Future studies exploring the variable/static aspects of interictal iEEG as well as if brain states or vigilance states affect the predictive value of these functional networks are needed.

“Taken together, our results support the use of interictal iEEG networks for predicting surgical outcomes and provide considerations and practical solutions for its clinical use,” Taylor and colleagues concluded. “Future studies should investigate the generalizability of the approach across multiple clinical sites and assess the combined use with other noninvasive whole‐brain modalities. The principles investigated here may also serve as an inspiration for the investigation of other neurological disorders.”

REFERENCE

Wang Y, Sinha N, Schroeder GM, et al. Interictal intracranial electroencephalography for predicting surgical success: The importance of space and time. Epilepsia. Published online June 26, 2020. doi: 10.1111/epi.16580