Current Literature In Clinical Science
Outcomes of Epilepsy Surgery for Epileptic Networks
Predicting Neurosurgical Outcomes in Focal Epilepsy Patients Using Computational Modelling. Sinha N, Dauwels J, Kaiser M, Cash SS, Brandon Westover M, Wang Y, Taylor PN. Brain 2017;140:319–332.
Surgery can be a last resort for patients with intractable, medically refractory epilepsy. For many of these patients, however, there is substantial risk that the surgery will be ineffective. The prediction of who is likely to benefit from a surgical approach is crucial for being able to inform patients better, conduct principled prospective clinical trials, and ultimately tailor therapeutic approaches to these patients more effectively. Dynamical computational models, informed with patient data, can be used to make predictions and give mechanistic insight. In this study, we develop patient-specific dynamical network models of epileptogenic cortex. We infer the network connectivity matrix from non-seizure electrographic recordings of patients and use these connectivity matrices as the network structure in our model. The model simulates the dynamics of a bi-stable switch at every node in this network, meaning that every node starts in a background state, but has the ability to transit to a co-existing seizure state. Whether a transition happens in a node is partly determined by the stochastic nature of the input to the node, but also by the input the node receives from other connected nodes in the network. By conducting simulations with such a model, we can detect the average transition time for nodes in a given network, and therefore define nodes with a short transition time as highly epileptogenic. In a retrospective study, we found that in some patients the regions with high epileptogenicity in the model overlap with those identified clinically as the seizure onset zone. Moreover, it was found that the resection of these regions in the model reduces the overall likelihood of a seizure. Following removal of these regions in the model, we predicted surgical outcomes and compared these to actual patient outcomes. Our predictions were found to be 81.3% accurate on a dataset of 16 patients with intractable epilepsy. Intriguingly, in patients with unsuccessful outcomes, the proposed computational approach is able to suggest alternative resection sites. The model presented here gives mechanistic insight as to why surgery may be unsuccessful in some patients. This may aid clinicians in presurgical evaluation by providing a tool to explore various surgical options, offering complementary information to existing clinical techniques.
Commentary Resective brain surgery to treat drug-resistant focal epilepsy is a major endeavor. Patients, family members, neurologists, neuroradiologists, neuropsychologists, neurosurgeons, and psychiatrists invest significant time and resources to localize the epilepsy and develop a safe and rational surgical strategy. The typical presurgical workup is usually extremely thorough with multiple electroencephalographic recordings, several sophisticated neuroimaging studies, and sometimes risky intracranial evaluations. Yet, the final critical step of determining the “epileptogenic zone” (EZ), that is, the “area of cortex that is necessary and sufficient for initiating seizures and whose removal (or disconnection) is necessary for complete abolition of seizures,” remains essentially a subjective assessment, relying heavily on the healthcare provider’s interpretation of fairly complex presurgical data points. Efforts attempting to infuse more objectivity into the definition of the EZ, such as the article highlighted in this commentary, are highly welcome. Epilepsy Currents, Vol. 17, No. 3 (May/June) 2017 pp. 160–162 © American Epilepsy Society
160
In this study, Sinha et al. report on dynamical computational models, informed with patient EEG data, that can “estimate” the extent of epileptogenicity in distinct brain regions and thus predict seizure outcome after epilepsy surgery should these regions be removed. This work falls in line with multiple other recent publications elaborating on either neuroimaging or electrophysiological representations of focal epilepsy as a network disease resulting from dysfunctional electrical connections across distinct brain regions, sometimes with a clear imaging correlate (atrophy of various hypothetically “connected” brain regions (1–3) or abnormal functional connectivity measures [4]) and sometimes supported by abnormal electrical connectivity patterns (5). The strengths of such a thought process are many. First, epilepsy surgery is conceptualized as a treatment aiming to remove a critical node in an epileptic network rather than an extraction of a limited epileptic focus. The mere appreciation of epilepsy as resulting from multiple connected brain regions, all with a potential for triggering seizures, albeit to varying degrees, facilitates the understanding of surgical failure as sometimes resulting from an incomplete disruption of this epileptic network rather than from mislocalization of the epileptic focus. Second, work attempting to let the data speak for itself through computational modeling
Epilepsy Surgery for Networks
algorithms, such as those used in Sinha et al., or through other methodologies, such as partial directed coherence (6) or effective connectivity analysis (7), elevates the exercise of localizing the EZ from the expertise based—essentially arbitrary—current practice to a more objective data-driven process. In an era where such advanced statistical tools are available, we have no excuse to do anything but embrace the potential of applying this statistical progress to improve our understanding of epilepsy and its various available treatments, including surgery. After all, surgical outcomes haven’t been revolutionized by our traditional “let’s find the lesion” efforts, so we don’t have much to lose by exploring statistics in the context of network definition. With all the promise of these novel approaches and concepts, where do we go? What are the caveats to remember? First, no data analysis can lead to conclusions that are more valid than what the quality of the data can handle. Specifically, in the context of analyzing invasive EEG data, as was done with Sinha et al. and all the other studies referenced so far, no amount of modeling—regardless of how sophisticated and advanced it is—can make up for the fact that invasive EEG recordings only inform about the limited region of cortex they are implanted in: subdural electrode recordings are blind to depths of sulci or deep regions, such as insula and cingulate cortex, while stereo-EEG electrodes leave much of the cortical surface unmeasured, thereby challenging a comfortable delineation of the cortical extent of epileptogenicity. Predictive analytics incorporating multiple testing modalities can provide an outlet optimizing the use of all knowledge collected in the course of a presurgical evaluation, rather than the inflated reliance on a single test. Second, efforts at improving outcome prediction in cohorts of epilepsy surgery patients should be paired with efforts at evolving better tools for individualized outcome prediction. The bottom line is that, as clinicians, our responsibility is to care for one patient at a time. Our duty is to translate the mass knowledge into meaningful guidance for how to manage the one patient who is waiting for our decision on how best to proceed with his or her care (8, 9). We should then get past the satisfaction of finding significant correlations between patterns from one or more tests in patient cohorts and surgical outcomes to an appreciation of the unmet need for a personalized medicine concept that includes personalized prognostication in addition to personalized diagnosis and treatment. Tools such as nomograms can start to fill this void as they become more fine-tuned (10). Third, epilepsy should not be expected to be restricted by time any more than it is expected to be restricted by space (11). As evidenced by this study of Sinha et al. and several others, we are now perceiving innovation in the idea that focal epilepsy is a network disease, while in fact, it has long been known that the whole brain is connected and that seizures cannot manifest unless multiple regions in this brain are involved in an abnormal electrical synchronization. The step forward is in the translation of this concept to surgical epilepsy as we propose that surgery fails sometimes because the “network is more extensive than the resection.” While that is certainly true in some cases, this logic still misses a critical mark. If we are embracing a “no space limitation” concept as
revolutionary and hypothesizing that the epileptic network is made of multiple nodes of varying degrees of epileptogenicity distributed in space across an epileptic brain, then why should we reject a “no time limitation” concept by expecting the epileptogenicity in these various nodes to somehow become static after surgery? Since we got ourselves to the point of accepting that focal epilepsy lives in multiple brain regions after a period of maturation into an epileptogenic complex structure (a network), then why not expand this network to the brain itself, to the very genetic molecular substrate that matured the current epileptic network and could continue to evolve the remaining nodes of “currently lower epileptogenicity” in the future? Epilepsy is thus equally likely to recur from either an incomplete or a nonsustained disruption of the epileptic network. In other words, sometimes seizures recur after surgery because the “network is more extensive than the resection,” but it is also possible that seizures recur because a “new network” is bound to develop in a given brain. Improving surgical outcomes should go beyond improving epilepsy localization. The challenge is surmountable: we just have to think a little bit differently, and studies like the one highlighted here get us a step closer. By Lara Jehi, MD References 1. Keller SS, Richardson MP, Schoene-Bake JC, O’Muircheartaigh J, Elkommos S, Kreilkamp B, Goh YY, Marson SG, Elger C. Weber B. Thalamotemporal alteration and postoperative seizures in temporal lobe epilepsy. Ann Neurol. 2015;77:760–774. 2. Munsell BC, Wee CY, Keller SS, Weber B, Elger C, da Silva LA, Nesland T, Styner M, Shen D, Bonilha L. Evaluation of machine learning algorithms for treatment outcome prediction in patients with epilepsy based on structural connectome data. Neuroimage. 2015;118:219– 230. 3. Bonilha L, Keller SS. Quantitative MRI in refractory temporal lobe epilepsy: Relationship with surgical outcomes. Quant Imaging Med Surg. 2015;5:204–224. 4. Jones SE, Zhang M, Avitsian R, Bhattacharyya P, Bulacio J, Cendes F, Enatsu R, Lowe M, Najm I, Nair D, Phillips M, Gonzalez-Martinez J. Functional magnetic resonance imaging networks induced by intracranial stimulation may help defining the epileptogenic zone. Brain Connect. 2014;4:286–298. 5. Jirsa VK, Proix T, Perdikis D, Woodman MM, Wang H, Gonzalez-Martinez J, Bernard C, Bénar C, Guye M, Chauvel P, Bartolomei F. The virtual epileptic patient: Individualized whole-brain models of epilepsy spread. Neuroimage. 2017;145:377–388. 6. Panzica F, Varotto G, Rotondi F, Spreafico R, Franceschetti S. Identification of the epileptogenic zone from stereo-EEG signals: A connectivity-graph theory approach. Front Neurol. 2013;4:175. 7. van Mierlo P, Carrette E, Hallez H, Raedt R, Meurs A, Vandenberghe S, Van Roost D, Boon P, Staelens S, Vonck K. Ictal-onset localization through connectivity analysis of intracranial EEG signals in patients with refractory epilepsy. Epilepsia. 2013;54:1409–1418. 8. Garcia Gracia C, Yardi R, Kattan MW, Nair D, Gupta A, Najm I, Bingaman W, Gonzalez-Martinez J, Jehi L. Seizure freedom score: A new simple method to predict success of epilepsy surgery. Epilepsia. 2015;56:359–365.
161
Epilepsy Surgery for Networks
9. Jehi L, Yardi R, Chagin K, Tassi L, Lo Russo G, Worrell G, Hu W, Cendes F, Morita M, Bartolomei f, Chauvel P, Najm I, Gonzalez-Martinez J, Bingaman W, Kattan MW. Development and validation of nomograms to provide individualised predictions of seizure outcomes after epilepsy surgery: A retrospective analysis. Lancet Neurol. 2015;14:283–290.
162
10. Jehi L, Chagin K, Yardi R, Tassi L, Russo GL, Worrell G, Cendes F, Morita M, Bartolomei F, Chauvel P, Najm I, Gonzalez-Martinez J, Bingaman W, Kattan MW. Promise and pitfalls of prognostic models for epilepsy surgery-Authors' reply. Lancet Neurol. 2015;14:684. 11. Jehi L. Improving seizure outcomes after epilepsy surgery: Time to break the “find and cut” mold. Epilepsy Curr. 2015;15:189–191.
Outcomes of Epilepsy Surgery for Epileptic Networks
Predicting Neurosurgical Outcomes in Focal Epilepsy Patients Using Computational Modelling. Sinha N, Dauwels J, Kaiser M, Cash SS, Brandon Westover M, Wang Y, Taylor PN. Brain 2017;140:319–332.
Surgery can be a last resort for patients with intractable, medically refractory epilepsy. For many of these patients, however, there is substantial risk that the surgery will be ineffective. The prediction of who is likely to benefit from a surgical approach is crucial for being able to inform patients better, conduct principled prospective clinical trials, and ultimately tailor therapeutic approaches to these patients more effectively. Dynamical computational models, informed with patient data, can be used to make predictions and give mechanistic insight. In this study, we develop patient-specific dynamical network models of epileptogenic cortex. We infer the network connectivity matrix from non-seizure electrographic recordings of patients and use these connectivity matrices as the network structure in our model. The model simulates the dynamics of a bi-stable switch at every node in this network, meaning that every node starts in a background state, but has the ability to transit to a co-existing seizure state. Whether a transition happens in a node is partly determined by the stochastic nature of the input to the node, but also by the input the node receives from other connected nodes in the network. By conducting simulations with such a model, we can detect the average transition time for nodes in a given network, and therefore define nodes with a short transition time as highly epileptogenic. In a retrospective study, we found that in some patients the regions with high epileptogenicity in the model overlap with those identified clinically as the seizure onset zone. Moreover, it was found that the resection of these regions in the model reduces the overall likelihood of a seizure. Following removal of these regions in the model, we predicted surgical outcomes and compared these to actual patient outcomes. Our predictions were found to be 81.3% accurate on a dataset of 16 patients with intractable epilepsy. Intriguingly, in patients with unsuccessful outcomes, the proposed computational approach is able to suggest alternative resection sites. The model presented here gives mechanistic insight as to why surgery may be unsuccessful in some patients. This may aid clinicians in presurgical evaluation by providing a tool to explore various surgical options, offering complementary information to existing clinical techniques.
Commentary Resective brain surgery to treat drug-resistant focal epilepsy is a major endeavor. Patients, family members, neurologists, neuroradiologists, neuropsychologists, neurosurgeons, and psychiatrists invest significant time and resources to localize the epilepsy and develop a safe and rational surgical strategy. The typical presurgical workup is usually extremely thorough with multiple electroencephalographic recordings, several sophisticated neuroimaging studies, and sometimes risky intracranial evaluations. Yet, the final critical step of determining the “epileptogenic zone” (EZ), that is, the “area of cortex that is necessary and sufficient for initiating seizures and whose removal (or disconnection) is necessary for complete abolition of seizures,” remains essentially a subjective assessment, relying heavily on the healthcare provider’s interpretation of fairly complex presurgical data points. Efforts attempting to infuse more objectivity into the definition of the EZ, such as the article highlighted in this commentary, are highly welcome. Epilepsy Currents, Vol. 17, No. 3 (May/June) 2017 pp. 160–162 © American Epilepsy Society
160
In this study, Sinha et al. report on dynamical computational models, informed with patient EEG data, that can “estimate” the extent of epileptogenicity in distinct brain regions and thus predict seizure outcome after epilepsy surgery should these regions be removed. This work falls in line with multiple other recent publications elaborating on either neuroimaging or electrophysiological representations of focal epilepsy as a network disease resulting from dysfunctional electrical connections across distinct brain regions, sometimes with a clear imaging correlate (atrophy of various hypothetically “connected” brain regions (1–3) or abnormal functional connectivity measures [4]) and sometimes supported by abnormal electrical connectivity patterns (5). The strengths of such a thought process are many. First, epilepsy surgery is conceptualized as a treatment aiming to remove a critical node in an epileptic network rather than an extraction of a limited epileptic focus. The mere appreciation of epilepsy as resulting from multiple connected brain regions, all with a potential for triggering seizures, albeit to varying degrees, facilitates the understanding of surgical failure as sometimes resulting from an incomplete disruption of this epileptic network rather than from mislocalization of the epileptic focus. Second, work attempting to let the data speak for itself through computational modeling
Epilepsy Surgery for Networks
algorithms, such as those used in Sinha et al., or through other methodologies, such as partial directed coherence (6) or effective connectivity analysis (7), elevates the exercise of localizing the EZ from the expertise based—essentially arbitrary—current practice to a more objective data-driven process. In an era where such advanced statistical tools are available, we have no excuse to do anything but embrace the potential of applying this statistical progress to improve our understanding of epilepsy and its various available treatments, including surgery. After all, surgical outcomes haven’t been revolutionized by our traditional “let’s find the lesion” efforts, so we don’t have much to lose by exploring statistics in the context of network definition. With all the promise of these novel approaches and concepts, where do we go? What are the caveats to remember? First, no data analysis can lead to conclusions that are more valid than what the quality of the data can handle. Specifically, in the context of analyzing invasive EEG data, as was done with Sinha et al. and all the other studies referenced so far, no amount of modeling—regardless of how sophisticated and advanced it is—can make up for the fact that invasive EEG recordings only inform about the limited region of cortex they are implanted in: subdural electrode recordings are blind to depths of sulci or deep regions, such as insula and cingulate cortex, while stereo-EEG electrodes leave much of the cortical surface unmeasured, thereby challenging a comfortable delineation of the cortical extent of epileptogenicity. Predictive analytics incorporating multiple testing modalities can provide an outlet optimizing the use of all knowledge collected in the course of a presurgical evaluation, rather than the inflated reliance on a single test. Second, efforts at improving outcome prediction in cohorts of epilepsy surgery patients should be paired with efforts at evolving better tools for individualized outcome prediction. The bottom line is that, as clinicians, our responsibility is to care for one patient at a time. Our duty is to translate the mass knowledge into meaningful guidance for how to manage the one patient who is waiting for our decision on how best to proceed with his or her care (8, 9). We should then get past the satisfaction of finding significant correlations between patterns from one or more tests in patient cohorts and surgical outcomes to an appreciation of the unmet need for a personalized medicine concept that includes personalized prognostication in addition to personalized diagnosis and treatment. Tools such as nomograms can start to fill this void as they become more fine-tuned (10). Third, epilepsy should not be expected to be restricted by time any more than it is expected to be restricted by space (11). As evidenced by this study of Sinha et al. and several others, we are now perceiving innovation in the idea that focal epilepsy is a network disease, while in fact, it has long been known that the whole brain is connected and that seizures cannot manifest unless multiple regions in this brain are involved in an abnormal electrical synchronization. The step forward is in the translation of this concept to surgical epilepsy as we propose that surgery fails sometimes because the “network is more extensive than the resection.” While that is certainly true in some cases, this logic still misses a critical mark. If we are embracing a “no space limitation” concept as
revolutionary and hypothesizing that the epileptic network is made of multiple nodes of varying degrees of epileptogenicity distributed in space across an epileptic brain, then why should we reject a “no time limitation” concept by expecting the epileptogenicity in these various nodes to somehow become static after surgery? Since we got ourselves to the point of accepting that focal epilepsy lives in multiple brain regions after a period of maturation into an epileptogenic complex structure (a network), then why not expand this network to the brain itself, to the very genetic molecular substrate that matured the current epileptic network and could continue to evolve the remaining nodes of “currently lower epileptogenicity” in the future? Epilepsy is thus equally likely to recur from either an incomplete or a nonsustained disruption of the epileptic network. In other words, sometimes seizures recur after surgery because the “network is more extensive than the resection,” but it is also possible that seizures recur because a “new network” is bound to develop in a given brain. Improving surgical outcomes should go beyond improving epilepsy localization. The challenge is surmountable: we just have to think a little bit differently, and studies like the one highlighted here get us a step closer. By Lara Jehi, MD References 1. Keller SS, Richardson MP, Schoene-Bake JC, O’Muircheartaigh J, Elkommos S, Kreilkamp B, Goh YY, Marson SG, Elger C. Weber B. Thalamotemporal alteration and postoperative seizures in temporal lobe epilepsy. Ann Neurol. 2015;77:760–774. 2. Munsell BC, Wee CY, Keller SS, Weber B, Elger C, da Silva LA, Nesland T, Styner M, Shen D, Bonilha L. Evaluation of machine learning algorithms for treatment outcome prediction in patients with epilepsy based on structural connectome data. Neuroimage. 2015;118:219– 230. 3. Bonilha L, Keller SS. Quantitative MRI in refractory temporal lobe epilepsy: Relationship with surgical outcomes. Quant Imaging Med Surg. 2015;5:204–224. 4. Jones SE, Zhang M, Avitsian R, Bhattacharyya P, Bulacio J, Cendes F, Enatsu R, Lowe M, Najm I, Nair D, Phillips M, Gonzalez-Martinez J. Functional magnetic resonance imaging networks induced by intracranial stimulation may help defining the epileptogenic zone. Brain Connect. 2014;4:286–298. 5. Jirsa VK, Proix T, Perdikis D, Woodman MM, Wang H, Gonzalez-Martinez J, Bernard C, Bénar C, Guye M, Chauvel P, Bartolomei F. The virtual epileptic patient: Individualized whole-brain models of epilepsy spread. Neuroimage. 2017;145:377–388. 6. Panzica F, Varotto G, Rotondi F, Spreafico R, Franceschetti S. Identification of the epileptogenic zone from stereo-EEG signals: A connectivity-graph theory approach. Front Neurol. 2013;4:175. 7. van Mierlo P, Carrette E, Hallez H, Raedt R, Meurs A, Vandenberghe S, Van Roost D, Boon P, Staelens S, Vonck K. Ictal-onset localization through connectivity analysis of intracranial EEG signals in patients with refractory epilepsy. Epilepsia. 2013;54:1409–1418. 8. Garcia Gracia C, Yardi R, Kattan MW, Nair D, Gupta A, Najm I, Bingaman W, Gonzalez-Martinez J, Jehi L. Seizure freedom score: A new simple method to predict success of epilepsy surgery. Epilepsia. 2015;56:359–365.
161
Epilepsy Surgery for Networks
9. Jehi L, Yardi R, Chagin K, Tassi L, Lo Russo G, Worrell G, Hu W, Cendes F, Morita M, Bartolomei f, Chauvel P, Najm I, Gonzalez-Martinez J, Bingaman W, Kattan MW. Development and validation of nomograms to provide individualised predictions of seizure outcomes after epilepsy surgery: A retrospective analysis. Lancet Neurol. 2015;14:283–290.
162
10. Jehi L, Chagin K, Yardi R, Tassi L, Russo GL, Worrell G, Cendes F, Morita M, Bartolomei F, Chauvel P, Najm I, Gonzalez-Martinez J, Bingaman W, Kattan MW. Promise and pitfalls of prognostic models for epilepsy surgery-Authors' reply. Lancet Neurol. 2015;14:684. 11. Jehi L. Improving seizure outcomes after epilepsy surgery: Time to break the “find and cut” mold. Epilepsy Curr. 2015;15:189–191.
