Intracranial Depth Electrodes Research Articles

Internal states as a source of subject-dependent movement variability are represented by large-scale brain networks

Humans’ ability to adapt and learn relies on reflecting on past performance. These experiences form latent representations called internal states that induce movement variability that improves how we interact with our environment. Our study uncovered temporal dynamics and neural substrates of two states from ten subjects implanted with intracranial depth electrodes while they performed a goal-directed motor task with physical perturbations. We identified two internal states using state-space models: one tracking past errors and the other past perturbations. These states influenced reaction times and speed errors, revealing how subjects strategize from trial history. Using local field potentials from over 100 brain regions, we found large-scale brain networks such as the dorsal attention and default mode network modulate visuospatial attention based on recent performance and environmental feedback. Notably, these networks were more prominent in higher-performing subjects, emphasizing their role in improving motor performance by regulating movement variability through internal states.

Ultraflexible endovascular probes for brain recording through micrometer-scale vasculature

Implantable neuroelectronic interfaces have enabled advances in both fundamental research and treatment of neurological diseases but traditional intracranial depth electrodes require invasive surgery to place and can disrupt neural networks during implantation. We developed an ultrasmall and flexible endovascular neural probe that can be implanted into sub-100-micrometer–scale blood vessels in the brains of rodents without damaging the brain or vasculature. In vivo electrophysiology recording of local field potentials and single-unit spikes have been selectively achieved in the cortex and olfactory bulb. Histology analysis of the tissue interface showed minimal immune response and long-term stability. This platform technology can be readily extended as both research tools and medical devices for the detection and intervention of neurological diseases.

Metabolic Changes in Medial Temporal Lobe Epilepsy Compared to Healthy Controls Using Advanced [18F] FDG PET SPM Analysis Techniques

Objectives: [18F] FDG PET imaging has emerged as an established modality in the evaluation of medically refractory temporal lobe epilepsy. Advanced techniques improve the sensitivity and detection of abnormalities. We sought to measure changes in metabolic activity using [18F] FDG PET in Medial Temporal Lobe Epilepsy (MTLE) patients compared to healthy controls and evaluate changes in the Seizure Onset Zone (SOZ) and remote areas. Methods: We evaluated a cohort of 14 MTLE patients. Following a standard brain [18F] FDG PET acquisition, images were co-registered to a healthy control database using the Neuro-MIM software. Z scores were generated for different temporal and extra-temporal lobe structures for significant p values <0.05. Seizure laterality was determined by experienced epileptologists (>10 years) using intracranial or surface depth electrodes. Results: The mean age of our patient cohort was 33.9 years (range: 9-52 years). We evaluated 7 left MTLE, 1 bilateral MTLE, and 6 right MTLE patients. Significant hypometabolic changes were seen in the amygdala, the hippocampus, and overall, the medial temporal lobe region (Z score of -2.2, -2.3, and -2.1 respectively). There was a smaller decrease in metabolic activity observed in the parahippocampal gyrus, fusiform gyrus, and the lateral temporal lobe (Z score of -1.4, -1.2, and -1.1 respectively). There were no significant differences in the basal ganglia, thalamus, prefrontal cortex, or cingulate gyrus. Conclusions: Measurable significant differences in FDG brain metabolic activity exist in medial temporal lobe epilepsy patients compared to healthy controls. Our study shows the most significant changes are hypometabolism in the amygdala, hippocampus, and overall, the medial temporal lobe region compared to the lateral temporal lobe.

835 Rapid Prototyping, Digital Sculpting, and Computer Aided Design for Point of Care 3D Printed Stereo EEG Models

INTRODUCTION: Part of the workup for surgical management of epilepsy is stereotactic placement of intracranial depth electrodes to better understand the origin and network pattern of spread of seizures. To facilitate surgical decision making, epilepsy team discussion, and patient/family communication, a workflow was developed to create 3D printed models of sEEG patient brains and electrode locations which could be invaluable in explaining EEG results and surgical options. METHODS: Pre-op MRI and post-implantation CT scans of a sEEG case were imported and merged using 3D Slicer, a free and open-source software package for advanced medical imaging research.The CT was segmented to include only the high-density metallic electrodes. These 3D models were exported into Autodesk Fusion 360, a computer-aided design (CAD) program, where electrode models were designed based on their location on the CT scan. These models were exported to Maxon Zbrush to be cleaned and re-meshed to make the model easier to sculpt and manipulate.The electrode objects were used as cutting tools to create holes in the brain model. Finally, the relevant portion of the brain (right temporal lobe and surrounding regions) and electrodes were exported for 3D printing. The final models were 3D printed on a FormLabs Form 3 stereolithography (SLA) printer using clear resin for the brain and flexible grey resin for the electrodes. RESULTS: CONCLUSIONS: A workflow was devised to print anatomic models of brains implanted with electrodes for seizure monitoring in 2-3 days utilizing software from multiple disciplines including medical imaging, CAD/engineering, and 3D sculpting used in the film and video game industries. Future models will include further anatomical definition with more advanced 3D sculpting and printing techniques.

Efficient volume-based localization and automatic labeling of intracranial depth electrodes.

The accurate localization and anatomical labeling of intracranial depth electrodes are crucial for stereoelectroencephalography (SEEG) recordings and the interpretation of results in patients with epilepsy. The laborious electrode localization procedure requires an efficient and easy-to-use pipeline. Thus, we developed a useful tool, which we called the depth electrode localizer (DELLO), to automatically identify and label depth electrode contacts with ease. The DELLO is an open-source package developed in MATLAB (MathWorks). It was specifically fine-tuned to expedite the localization of depth electrodes. The basic procedures include preoperative magnetic resonance imaging (MRI) and postoperative computed tomography coregistration, intensity threshold electrode spatial sampling, the hierarchical clustering of electrode samples, and gray-matter and automatic anatomical labeling (AAL). The DELLO also has a graphical user interface (GUI) that can be used to review the results. The only manual intervention procedures are the identification of the target (tip) and entry point of each electrode and the naming of the clustered electrode contact groups, which generally take ~5 min per case. The coordinates of each contact were recorded in individual spaces and were also transformed in standard space by applying a volume-based deformation field. To validate the performance of the current method, 7 patients with epilepsy were retrospectively included in the analysis. A total of 80 depth electrodes, including 1,030 contacts from the 7 patients with epilepsy, were localized. All the procedures functioned well, and the entire process was robust and intuitive. Among the 1,030 contacts, 746 (72.43%) were labeled as inside the gray matter. The gray-matter and AAL accuracy rates were 95.83% and 90.78%, respectively, over all contacts. The DELLO is an integrated tool that was designed to semi-automatically localize and label intracranial depth electrodes. It is open source and freely available. Given its high accuracy and efficiency, the DELLO could facilitate SEEG interpretation and be used in SEEG-based cognitive neuroscience studies.

Synthesizing Speech from Intracranial Depth Electrodes using an Encoder-Decoder Framework

Speech Neuroprostheses have the potential to enable communication for people with dysarthria or anarthria. Recent advances have demonstrated high-quality text decoding and speech synthesis from electrocorticographic grids placed on the cortical surface. Here, we investigate a less invasive measurement modality in three participants, namely stereotactic EEG (sEEG) that provides sparse sampling from multiple brain regions, including subcortical regions. To evaluate whether sEEG can also be used to synthesize high-quality audio from neural recordings, we employ a recurrent encoder-decoder model based on modern deep learning methods. We find that speech can indeed be reconstructed with correlations up to 0.8 from these minimally invasive recordings, despite limited amounts of training data.

Modeling intracranial electrodes. A simulation platform for the evaluation of localization algorithms.

IntroductionIntracranial electrodes are implanted in patients with drug-resistant epilepsy as part of their pre-surgical evaluation. This allows the investigation of normal and pathological brain functions with excellent spatial and temporal resolution. The spatial resolution relies on methods that precisely localize the implanted electrodes in the cerebral cortex, which is critical for drawing valid inferences about the anatomical localization of brain function. Multiple methods have been developed to localize the electrodes, mainly relying on pre-implantation MRI and post-implantation computer tomography (CT) images. However, they are hard to validate because there is no ground truth data to test them and there is no standard approach to systematically quantify their performance. In other words, their validation lacks standardization. Our work aimed to model intracranial electrode arrays and simulate realistic implantation scenarios, thereby providing localization algorithms with new ways to evaluate and optimize their performance.ResultsWe implemented novel methods to model the coordinates of implanted grids, strips, and depth electrodes, as well as the CT artifacts produced by these. We successfully modeled realistic implantation scenarios, including different sizes, inter-electrode distances, and brain areas. In total, ∼3,300 grids and strips were fitted over the brain surface, and ∼850 depth electrode arrays penetrating the cortical tissue were modeled. Realistic CT artifacts were simulated at the electrode locations under 12 different noise levels. Altogether, ∼50,000 thresholded CT artifact arrays were simulated in these scenarios, and validated with real data from 17 patients regarding the coordinates’ spatial deformation, and the CT artifacts’ shape, intensity distribution, and noise level. Finally, we provide an example of how the simulation platform is used to characterize the performance of two cluster-based localization methods.ConclusionWe successfully developed the first platform to model implanted intracranial grids, strips, and depth electrodes and realistically simulate thresholded CT artifacts and their noise. These methods provide a basis for developing more complex models, while simulations allow systematic evaluation of the performance of electrode localization techniques. The methods described in this article, and the results obtained from the simulations, are freely available via open repositories. A graphical user interface implementation is also accessible via the open-source iElectrodes toolbox.

Interhemispheric theta coherence in the hippocampus for successful object-location memory in human intracranial encephalography

Our knowledge concerning visual-spatial memory related phase synchronization within the ipsilateral hippocampus or between contralateral hippocampi during memory encoding in humans is currently limited. The present study examines the relationship between phase synchronization within the hippocampus and memory performance during virtual navigation in an object-location memory navigation task using intracranial depth electrodes in human subjects. Specifically, we focus on the phase synchronization ratio between periods when the target object was in and out of visual focus. Our findings indicate that there is a significant relationship between this phase synchronization ratio and object-location memory performance in the theta band (p = 0.015, R = −0.71), but not in the delta or alpha bands. Importantly, this theta coherence has a significant linear relationship with memory performance between contralateral hippocampus electrode pairs (p = 0.006, R = −0.77), but not ipsilateral electrode pairs (p = 0.79, R = −0.09). In addition, this theta coherence has a significant linear relationship with memory performance during stationary periods (p = 0.002, R = −0.82), but not movement periods (p = 0.10, R = −0.51). These findings suggest that, during navigation, interhemispheric hippocampal theta coherence when stationary and focusing on the target object may be a critical determinant of successful object-location memory.

0147 Electrophysiologic Localization of Pathological Brain Tissue Is Accomplished with Intracranial Sleep Staging

0147 Electrophysiologic Localization of Pathological Brain Tissue Is Accomplished with Intracranial Sleep Staging

Neuromodulation in Beta-Band Power Between Movement Execution and Inhibition in the Human Hippocampus

IntroductionThe hippocampus is thought to be involved in movement, but its precise role in movement execution and inhibition has not been well studied. Previous work with direct neural recordings has found beta-band (13–30 Hz) modulation in both movement execution and inhibition throughout the motor system, but the role of beta-band modulation in the hippocampus during movement inhibition is not well understood. Here, we perform a Go/No-Go reaching task in ten patients with medically refractory epilepsy to study human hippocampal beta-power changes during movement. Materials and MethodsTen epilepsy patients (5 female; ages 21–46) were implanted with intracranial depth electrodes for seizure monitoring and localization. Local field potentials were sampled at 2000 Hz during a Go/No-Go movement task. Comparison of beta-band power between Go and No-Go conditions was conducted using Wilcoxon signed-rank hypothesis testing for each patient. Sub-analyses were conducted to assess differences in the anterior vs posterior contacts, ipsilateral vs contralateral contacts, and male vs female beta-power values. ResultsEight out of ten patients showed significant beta-power decreases during the Go movement response (p < 0.05) compared to baseline. Eight out of ten patients also showed significant beta-power increases in the No-Go condition, occurring in the absence of movement. No significant differences were noted between ipsilateral vs contralateral contacts nor in anterior vs posterior hippocampal contacts. Female participants had a higher task success rate than males and had significantly greater beta-power increases in the No-Go condition (p < 0.001). ConclusionThese findings indicate that increases in hippocampal beta power are associated with movement inhibition. To the best of our knowledge, this study is the first to report this phenomenon in the human hippocampus. The beta band may represent a state-change signal involved in motor processing. Future focus on the beta band in understanding human motor and impulse control will be vital.

Amygdaloid theta-band power increases during conflict processing in humans

The amygdala is a medial temporal lobe structure known to be involved in processing emotional conflict. However, its role in processing non-emotional conflict is not well understood. Previous studies have utilized the Stroop Task to examine brain modulation of humans under the color-word conflict scenario, which is non-emotional conflict processing, and found hippocampal theta-band (4–7 Hz) modulation. This study aims to survey amygdaloid theta power changes during non-emotional conflict processing using intracranial depth electrodes in nine epileptic patients (3 female; age 20–62). All patients were asked to perform a modified Stroop task. During task performance, local field potential (LFP) data was recorded from macro contacts sampled at 2 K Hz and used for analysis. Mean theta power change from baseline was compared between the incongruent and congruent task condition groups using a paired sample t-test. Seven patients were available for analysis after artifact exclusion. In five out of seven patients, statistically significant increases in theta-band power from baseline were noted during the incongruent task condition (paired sample t-test p < 0.001), including one patient exhibiting theta power increases in both task conditions. Average response time was 1.07 s (failure trials) and 1.04 s (success trials). No speed-accuracy tradeoff was noted in this analysis. These findings indicate that human amygdaloid theta-band modulation may play a role in processing non-emotional conflict. It builds directly upon work suggesting that the amygdala processes emotional conflict and provides a neurophysiological mechanism for non-emotional conflict processing as well.

Generalizing neural signal-to-text brain-computer interfaces

Objective: Brain-Computer Interfaces (BCI) may help patients with faltering communication abilities due to neurodegenerative diseases produce text or speech by direct neural processing. However, their practical realization has proven difficult due to limitations in speed, accuracy, and generalizability of existing interfaces. The goal of this study is to evaluate the BCI performance of a robust speech decoding system that translates neural signals evoked by speech to a textual output. While previous studies have approached this problem by using neural signals to choose from a limited set of possible words, we employ a more general model that can type any word from a large corpus of English text. Approach: In this study, we create an end-to-end BCI that translates neural signals associated with overt speech into text output. Our decoding system first isolates frequency bands in the input depth-electrode signal encapsulating differential information regarding production of various phonemic classes. These bands form a feature set that then feeds into a Long Short-Term Memory (LSTM) model which discerns at each time point probability distributions across all phonemes uttered by a subject. Finally, a particle filtering algorithm temporally smooths these probabilities by incorporating prior knowledge of the English language to output text corresponding to the decoded word. The generalizability of our decoder is driven by the lack of a vocabulary constraint on this output word. Main result: This method was evaluated using a dataset of 6 neurosurgical patients implanted with intra-cranial depth electrodes to identify seizure foci for potential surgical treatment of epilepsy. We averaged 32% word accuracy and on the phoneme-level obtained 46% precision, 51% recall and 73.32% average phoneme error rate while also achieving significant increases in speed when compared to several other BCI approaches. Significance: Our study employs a more general neural signal-to-text model which could facilitate communication by patients in everyday environments.

Safety of MRI in the localization of implanted intracranial electrodes for refractory epilepsy.

This is an observational study to evaluate the safety of magnetic resonance imaging (MRI) to localize subdural grids and depth electrodes in patients with refractory epilepsy using a 1.5 Tesla MR scanner. We implemented an optimized MRI protocol providing adequate image quality for the assessment of subdural grids and depth electrodes, while minimizing the specific absorption rate (SAR). We reviewed all MRI studies performed in patients with subdural grids and depth electrodes between January 2010 and October 2018. Image quality was graded as acceptable or nonacceptable for the assessment of intracranial device positioning. We reviewed the medical record and any imaging obtained after intracranial implant removal for adverse event or complication occurring during and after the procedure. Ninety-nine patients with refractory epilepsy underwent MRI scans using a magnetization-prepared rapid acquisition of gradient echo sequence and a transmit-receive head coil with depth electrodes and subdural grids in place. Two patients underwent two separate depth electrode implantations for a total of 101 procedures and MRI scans. No clinical adverse events were reported during or immediately after imaging. Image quality was graded as acceptable for 97 MRI scans. Review of follow-up CT and MRI studies after implant removal, available for 70 patients, did not demonstrate unexpected complications in 69 patients. In our experience, a low SAR MRI protocol can be used to safely localize intracranial subdural grids and depth electrode in patients with refractory epilepsy.

Pediatric Epilepsy Surgery: Indications and Evaluation.

Epilepsy is a common neurological condition in children. It is usually amenable to drug therapy. However, nearly one-third of patients may be refractory to antiseizure drugs. Poor compliance and nonepileptic events should be ruled out as possible causes of drug-resistant epilepsy (DRE). After failing adequate trials of two appropriate antiseizure drugs, patients with focal DRE or poorly classifiable epilepsy or epileptic encephalopathy with focal electro-clinical features should be worked up for surgical candidacy. A randomized controlled trial provided a class I evidence for epilepsy surgery in pediatric DRE. Pre-surgical screening workup typically includes a high-resolution epilepsy protocol brain magnetic resonance imaging (MRI) and a high-quality in-patient video electroencephalography evaluation. Advanced investigations such as positron emission tomography (PET), single-photon emission computed tomography (SPECT), and magnetoencephalography (MEG) may be required in selected cases especially when brain MRI is normal, and further evidence for anatomo-electro-clinical concordance is necessary to refine candidacy for surgery and surgical strategy. Some children may also need functional MRI to map eloquent regions of interest such as motor, sensory, and language functions to avoid unacceptable neurological deficits after surgery. Selected children may need invasive long-term electroencephalographic monitoring using stereotactically implanted intracranial depth electrodes or subdural grids. Surgical options include resective surgeries (lesionectomy, lobectomy, multilobar resections) and disconnective surgeries (corpus callosotomy, etc.) with the potential to obtain seizure freedom. Other surgical procedures, typically considered to be palliative are neuromodulation [deep brain stimulation (DBS), vagal nerve stimulation (VNS), and responsive neural stimulation (RNS)]. DBS and RNS are currently not approved in children. Pediatric DRE should be evaluated early considering the risk of epileptic encephalopathy and negative impact on cognition.

Electrical stimulation of the medial orbitofrontal cortex in humans elicits pleasant olfactory perceptions

BackgroundOlfactory hallucinations can be part of epileptic seizures of orbitofrontal origin. Olfactory hallucinations, however, are rare and therefore the semiology, localization and lateralization characteristics are underdetermined. In addition, many discrepancies are found in the literature regarding olfactory processing and orbitofrontal (OF) functions and olfactory function. Particularly, the questions of laterality and affective component in coding of odors in the OF cortex remain controversial. AimsThis study explored whether cortical electrical stimulation of the OF and mesiotemporal brain can trigger olfactory hallucinations with special focus on olfactory percepts in terms of laterality and hedonics. Materials and methodsEight patients with temporal lobe epilepsy participated in the study, at the time of invasive exploration of their epilepsy. The most distal contact of the OF and anterior hippocampus depth electrodes were stimulated (50 Hz, 0.2 ms biphasic pulse; maximal stimulation 4 mA). Patients were instructed to report any kind of sensation they might experience. Intracranial depth electrodes were localized (iElectrodes): subject-specific brain mask, subcortical segmentation and cortical parcellation based on the Destrieux atlas (FreeSurfer) were superposed to the coregistered T1-weighted MRI and CT images (SPM). The center of mass of each electrode-artifact cluster determined the electrode localization. The electrode labeling was done in patient space. To obtain the electrode coordinates in Montreal Neurological Institute (MNI) space, the images obtained previously in the patient space were first segmented and normalized (SPM). Then, the localization procedure (iElectrodes) was run again with these new normalized images in MNI space. ResultsNo hallucination was evoked by stimulation, neither of the right nor the left hippocampus (8/8 patients). Pleasant olfactory hallucinations were evoked by OF stimulation in 5/8 patients in either hemisphere. Patients named the percept as the smell of lemon or coffee for example. Among those 5 patients, electrodes were localized in the cortex of the olfactory sulcus, medial orbital sulcus or medial OF gyrus. Increasing stimulation amplitude changed the olfactory percept identification in 3 out of those 5 patients. No affective judgement or change in perceived odor intensity was reported by the patients. No hallucination was evoked by the stimulation of the white matter of the medial OF brain in 3/8 patients independently of the hemisphere stimulated. ConclusionsThis study demonstrated that stimulation of the cortex of the medial OF brain and not of its white matter elicits specific pleasant olfactory hallucinations independently of the hemisphere stimulated, supporting one symmetrical olfactory processing in human.

Surgical Robot-Enhanced Implantation of Intracranial Depth Electrodes for Single Neuron Recording Studies in Patients with Medically Refractory Epilepsy: A Technical Note

Single neuron or unit recording enables researchers to measure the electrophysiologic responses of single neurons using a microelectrode system. This approach is widely used in cognitive science and has become more widespread in humans with the use of hybrid (micro-within-macrowire) depth electrodes that enable the implantation of microwires into the brain parenchyma. The authors describe their surgical technique in a total of 7 patients with intractable epilepsywho underwent robot-enhanced stereoencephalography in which both standard (nonhybrid) and hybrid depth electrodes were used for invasive chronic monitoring. The technique and accuracy of the procedure were evaluated with a total of 84 depth electrodes (46 hybrid, 38 standard) in 7 patients. No major complications, such as intracranial hemorrhage, infection or cerebrospinal fluid leakage, occurred regardless of the type of electrode used. The addition of hybrid depth electrodes for the purpose of invivo single neuron recording in robot-enhanced stereoencephalography procedures is safe and does not impact the accuracy of targeting or patient safety.

Combining 3D-printing Technology with Pre-operative Sterotactic Planning for Placement of Intracranial Depth Electrodes for Improved Patient Safety and Operative Efficiency: An Institutional Experience (2558)

Combining 3D-printing Technology with Pre-operative Sterotactic Planning for Placement of Intracranial Depth Electrodes for Improved Patient Safety and Operative Efficiency: An Institutional Experience (2558)

Sleep modulates effective connectivity: A study using intracranial stimulation and recording

ObjectiveSleep is an active process with an important role in memory. Epilepsy patients often display a disturbed sleep architecture, with consequences on cognition. We aimed to investigate the effect of sleep on cortical networks’ organization. MethodsWe analyzed cortico-cortical evoked responses elicited by single pulse electrical stimulation (SPES) using intracranial depth electrodes in 25 patients with drug-resistant focal epilepsy explored using stereo-EEG. We applied the SPES protocol during wakefulness and NREM – N2 sleep. We analyzed 31,710 significant responses elicited by 799 stimulations covering most brain structures, epileptogenic or non-epileptogenic. We analyzed effective connectivity between structures using a graph-theory approach. ResultsSleep increases excitability in the brain, regardless of epileptogenicity. Local and distant connections are differently modulated by sleep, depending on the tissue epileptogenicity.In non-epileptogenic areas, frontal lobe connectivity is enhanced during sleep. There is increased connectivity between the hippocampus and temporal neocortex, while perisylvian structures are disconnected from the temporal lobe. In epileptogenic areas, we found a clear interhemispheric difference, with decreased connectivity in the right hemisphere during sleep. ConclusionsSleep modulates brain excitability and reconfigures functional brain networks, depending on tissue epileptogenicity. SignificanceWe found specific patterns of information flow during sleep in physiologic and pathologic structures, with possible implications for cognition.

Ictal vomiting; A dominant hemisphere phenomenon as demonstrated by intracranial depth electrode seizure mapping

Ictal vomiting; A dominant hemisphere phenomenon as demonstrated by intracranial depth electrode seizure mapping

Automated unsupervised behavioral state classification using intracranial electrophysiology

Objective. Automated behavioral state classification in intracranial EEG (iEEG) recordings may be beneficial for iEEG interpretation and quantifying sleep patterns to enable behavioral state dependent neuromodulation therapy in next generation implantable brain stimulation devices. Here, we introduce a fully automated unsupervised framework to differentiate between awake (AW), sleep (N2), and slow wave sleep (N3) using intracranial EEG (iEEG) only and validated with expert scored polysomnography. Approach. Data from eight patients undergoing evaluation for epilepsy surgery (age , three female) with intracranial depth electrodes for iEEG monitoring were included. Spectral power features (0.1–235 Hz) spanning several frequency bands from a single electrode were used to classify behavioral states of patients into AW, N2, and N3. Main results. Overall, classification accuracy of 94%, with 94% sensitivity and 93% specificity across eight subjects using multiple spectral power features from a single electrode was achieved. Classification performance of N3 sleep was significantly better (95%, sensitivity 95%, specificity 93%) than that of the N2 sleep phase (87%, sensitivity 78%, specificity 96%). Significance. Automated, unsupervised, and robust classification of behavioral states based on iEEG data is possible, and it is feasible to incorporate these algorithms into future implantable devices with limited computational power, memory, and number of electrodes for brain monitoring and stimulation.