Hello, my name is Barnali Kundu and the title of my talk is a systematic exploration of parameters affecting evoked intracranial potentials in patients with epilepsy. I'd like to thank the AANS committee for giving me a chance to talk about our work. So there has been a movement from understanding epilepsy as a disease of a focus to the disease of a network. Numerous pieces of data show now that the epileptic brain has different topological features compared to a normal brain, and several fMRI studies to date have shown that the epileptic zone shows increased connectivity and that there's actually decreased connectivity in other global brain networks such as the default mode network. Important for surgery field, surgical resection has been related to abnormalities, both structural as well as functional in other areas of the brain, even contralateral areas. And so understanding these network features will not only help understand epilepsy as a disease, but also help isolate targets for surgery. CCEPs or corticocortical evoked potentials offers a means of quantifying these network features. So what are CCEPs? They are evoked potentials that are generated through stimulating macro electrodes, either surface or depth electrodes. They're on the order of 100 to 1,000 microvolts and last greater than 500 milliseconds. So this is what it looks like in the brain. This is stimulating the red starred electrode. These are depth electrodes. You can see the map of the voltage distribution across electrodes with one stimulation. And in the time domain, this is the same map with electrodes, stimulation delivered here in the red box, and you can see the distribution of voltages along time in the adjacent electrodes with the gray boxes is gray matter, white boxes is white matter, and the red box outline are the seizure onset zone. CCEPs are a measure of effective connectivity, which is the causal interaction between brain areas or networks. And that's different from structural or functional connectivity, structural being derived from, for example, DTI-based measures and functional, which is the correlation between two brain areas. At a cellular level, it's thought that CCEPs are generated at the axonal level over somas. And to date, there has been numerous pieces of evidence showing there's abnormalities in the CCEP when they are generated within the seizure onset zone. CCEP propagation has also been thought to reflect how seizures propagate. And importantly, resection of regions with abnormal CCEP features has been related to better seizure control. So CCEPs seem to be offering a great measure of effective connectivity. However, there are some issues with using them. There are several theoretical factors that affect the CCEP amplitude, including the stimulation current delivered, the electropolarity and tissue type. And unfortunately, there's not a lot of data out there describing how changing these stem parameters would affect the CCEP features. And unfortunately, that does limit interpretation of data. For example, from our group, we published a study not long ago where we were trying to identify the seizure onset zone based on CCEP features, and we're only able to do so in about half the patients. And part of the reason might be because we were stimulating only at 3 milliamps, and that might not have been optimal. So the objective of the current study was to better understand factors affecting the CCEP amplitude. We performed a prospective cohort study. We recruited adult patients who are undergoing long-term monitoring for seizures. We applied stimulation to a range of electrodes over a range of currents and recorded the evoked responses. We then did post-processing on that data to generate trial-by-trial CCEP amplitudes, as well as evoked gamma band power measurements, which I'll talk about later. Finally, and what is I think the strength of this study, is that we used a mixed effects model to characterize the CCEP amplitude as a function of multiple variables, including stimulation current, distance between stimulation and response channel, tissue type, and we're able to use patient as a random variable. Just briefly, the protocol for our stimulation to generate a CCEP, for each subject, we took their post-op CT with locations of the electrodes and merged it with their anatomical MRI scan. We then co-registered that with a standard atlas to be able to look at electrode locations, as well as whether or not they were in gray matter. For stimulation, we grounded 20 channels and used a white matter reference. We used the BlackRock system for both data acquisition as well as stimulation. Those are the stimulation parameters and data acquisition parameters. Importantly, we used the data from 0 to 100 milliseconds post-STIM, which is considered the early period. We used 11 patients in this study, and we had over 4,000 STIM response pairs, so we had a very large data set. We stimulated over a range of currents, 2.5 to 10 milliamps. This data is freely available on Mendeley data for anyone who wishes to analyze it themselves. The first parameter in our model was the stimulation current, and we found that, as expected, increasing stimulation current generates a larger CCEP. That is true across brain areas, patients, and tissue types. With increasing stimulation current, there's an increase in CCEP amplitude, but it does plateau at around 7.5 milliamps. We confirmed that with a separate data set where we stimulated at much smaller intervals and found that the CCEP amplitude plateaus around 5.5 to 7.5 milliamps. Based on the model, we can say there is about a 3% increase in CCEP amplitude per milliamp increase in stimulation. Also, there's a plateau effect at around 5.5 to 7.5 milliamps, and that may be related to the virtual anode effect when you're stimulating with cathodic stimulation that will restrict depulsation of axons. The second parameter in the model was the distance between STIM and response channel. We found the CCEP amplitude actually decreases as you increase the distance, as expected. In this case, this is a grid, and we're stimulating in the red box, and you can see that in the raw data, there's a decrease in the amplitude of the CCEP as you increase the distance away. That's true for the entire data set, and from the model, we can say there's a 3% decrease in CCEP amplitude per millimeter increase in distance. That's likely related to the fact that the potential generated from a point source does fall off with one of the distance. Another measure that we looked at was the evoked gamma band power generated by stimulation. That was motivated by the findings from the high-frequency oscillation literature, which show that HFOs, which look like this, they're basically very small blips of high-frequency activity within interictal data, are a marker of the seizure onset zone. Unfortunately, these HFOs can be notoriously hard to measure and detect. A group from Cleveland Clinic, actually, were able to generate this HFO phenomenon using stimulation, and within MTLE patients, they showed that stimulation, that evoked gamma band power generated from stimulation was related to location of the seizure onset zone. We wanted to automate that analysis and generalize across electrodes, and so we created a second model, a mixed-effects model with evoked gamma band power as the dependent variable, and looked at whether changing these other parameters would affect the amplitude of that power. We used the 70 to 150 hertz bandpass filter data because that is the most physiologically relevant and likened to neuronal activity. In this model, we found that increasing stimulation current, again, increases the evoked gamma band power, and that power are the black trials in this diagram. Over the entire dataset, we found that increasing the current increases the evoked gamma band power, but that it does plateau, similar to CCEP amplitude, around 7.5 milliamps, and that the power drops off with distance, so the distance relationship is slightly different, where there's basically a precipitous falloff in gamma band power after 3.5 centimeters. When looking across tissue types, we found that the CCEP amplitude and evoked gamma band power generated were largest within the hippocampus compared to neocortex, for example. This is true for both measures, and that might be related to the unique archicortex of hippocampus versus the other tissue types, and this phenomenon has been described in the literature. Lastly, but most importantly, we found that the CCEP amplitude and evoked gamma band power were highest in the seizure onset zone. That response was about 10% higher for the CCEP amplitude, and the evoked gamma band power was actually 120% higher in the seizure onset zone. This was an automated analysis, and so it can be incorporated into any clinician's mapping protocol. Several limitations to this study. One is that we were only able to have 10 trials per stim response pair per condition, and that there are also inherent inaccuracies in co-registration across imaging, which may lead to some minor mislabeling of electrodes if they're in border zones. There's also model limitations, where the model will not capture nonlinearities that aren't explicitly modeled. For example, there's a well-known phenomenon of CCEPs propagating to the contralateral hemisphere, and our model would not capture that as well, because most responses are generated very close to the electrode. So in conclusion, CCEP amplitude and evoked gamma band power are reliable measurements of effective connectivity, but they are variable across stim parameters and brain tissue types. For depth electrodes, at least we can say that stimulating at 5.5 milliamps or above will generate a maximal, reliable response, which would aid in comparison across patients and experimental conditions. Finally, the evoked gamma band power and CCEP amplitude are largest in the seizure onset zone compared to outside, and this automated protocol we provide could be incorporated into seizure mapping for a patient, potentially. In the future, we'd like to determine normative values for the CCEP amplitude and evoked gamma band power within and outside the seizure onset zone across tissue types. And of course, in our study, we didn't look at surgical outcomes, but it will be essential to determine if resecting areas with elevated CCEP amplitude or evoked gamma band power do indeed relate to better clinical outcomes. I'd like to acknowledge Dr. Caldwell and Dr. Jensen, as well as the department and our funding sources. These are the references. Please feel free to email me with any questions. Thank you.

evoked intracranial potentials

epilepsy

network features

CCEPs

effective connectivity