Prediction of Surgical Outcome by Analysis of Network Dynamics

Epilepsy surgery is the standard of care in eligible individuals with intractable epilepsy. Studies have shown that when seizures are pharmacoresistant (are not controlled with medication), resective surgery is way superior to continued medical treatment in terms of achieving seizure freedom.

Predicting Success

Numerous studies have attempted to identify predictors of surgical outcomes in temporal lobe epilepsy and frontal lobe epilepsy, among others.

• Lesional epilepsy (i.e., one that is related to a visible lesion in brain imaging studies) often improves surgical candidacy, as one common predictor of favorable outcome is inclusion of the epileptogenic lesion in the resection.
• Another predictor is resection of high-frequency-oscillation-producing tissue.
• Yet other predictors are related to identification of the extent of the irritability network. For example, seizures that have less tendency to propagate from one side of the brain to the other and ones that infrequently secondarily generalize tend to be associated with better surgical outcome.

Study Used Math to Predict Success

A recent paper suggested that applying mathematical models to estimate ictogenicity (likelihood of generating seizures) of a brain network can predict surgical outcome.¹ Networks consist of nodes and edges, which connect the nodes.

The authors used techniques arising from the complexity theory to investigate the relationship between node dynamics, patterns of connectivity within a network, and seizure generation in silico. They mathematically defined brain network ictogenicity (BNI) and node ictogenicity (NI) and applied these definitions to electrocorticograms of individuals who have undergone resection surgery with known surgical outcomes.

Study Conclusions

• They concluded that in those with good surgical outcome, their model prediction for optimal resection strategy tended to be consistent with the actual surgery.
• In those with poor surgical outcome, the model prediction and the surgical resection were different.

Interestingly, the authors claim that, contrary to previous studies, the best resection strategies are not determined by the locations of the epileptogenic lesion or the epileptic discharges. Although these claims may be surprising to clinicians who are aware of the epilepsy literature that established the predicting values of such imaging and EEG findings to the surgical outcome, the study is still important as it reminds us that there is more to be harvested from intracranial recordings than our habitual visual inspection with “naked eyes.”

It is important to note that model predictions are dependent on the implantation strategy. Obviously, if intracranial electrodes are far from the epileptogenic region, the outcome will not be favorable whether the model works or not. Therefore, it would be helpful if studying seizure dynamics can be done with scalp recordings even prior to, and perhaps informing, electrode implantation. An obvious problem with noninvasive recordings is the high propensity for recording muscle and environmental artifact, as well as the inverse problem and major signal filtering of the skull.

Reference

1. Goodfellow M, Rummel C, Abela E, Richardson MP, Schindler K, Terry JR. Estimation of brain network ictogenicity predicts outcome from epilepsy surgery. Sci Rep. 2016 Jul 7;6:29215.