Faculty Bibliography

OBJECTIVE/UNASSIGNED:This study investigates speech decoding from neural signals captured by intracranial electrodes. Most prior works can only work with electrodes on a 2D grid (i.e., Electrocorticographic or ECoG array) and data from a single patient. We aim to design a deep-learning model architecture that can accommodate both surface (ECoG) and depth (stereotactic EEG or sEEG) electrodes. The architecture should allow training on data from multiple participants with large variability in electrode placements and the trained model should perform well on participants unseen during training. APPROACH/UNASSIGNED:We propose a novel transformer-based model architecture named SwinTW that can work with arbitrarily positioned electrodes, by leveraging their 3D locations on the cortex rather than their positions on a 2D grid. We train both subject-specific models using data from a single participant as well as multi-patient models exploiting data from multiple participants. MAIN RESULTS/UNASSIGNED:The subject-specific models using only low-density 8x8 ECoG data achieved high decoding Pearson Correlation Coefficient with ground truth spectrogram (PCC=0.817), over N=43 participants, outperforming our prior convolutional ResNet model and the 3D Swin transformer model. Incorporating additional strip, depth, and grid electrodes available in each participant (N=39) led to further improvement (PCC=0.838). For participants with only sEEG electrodes (N=9), subject-specific models still enjoy comparable performance with an average PCC=0.798. The multi-subject models achieved high performance on unseen participants, with an average PCC=0.765 in leave-one-out cross-validation. SIGNIFICANCE/UNASSIGNED:The proposed SwinTW decoder enables future speech neuroprostheses to utilize any electrode placement that is clinically optimal or feasible for a particular participant, including using only depth electrodes, which are more routinely implanted in chronic neurosurgical procedures. Importantly, the generalizability of the multi-patient models suggests the exciting possibility of developing speech neuroprostheses for people with speech disability without relying on their own neural data for training, which is not always feasible.

Effective communication hinges on a mutual understanding of word meaning in different contexts. We recorded brain activity using electrocorticography during spontaneous, face-to-face conversations in five pairs of epilepsy patients. We developed a model-based coupling framework that aligns brain activity in both speaker and listener to a shared embedding space from a large language model (LLM). The context-sensitive LLM embeddings allow us to track the exchange of linguistic information, word by word, from one brain to another in natural conversations. Linguistic content emerges in the speaker's brain before word articulation and rapidly re-emerges in the listener's brain after word articulation. The contextual embeddings better capture word-by-word neural alignment between speaker and listener than syntactic and articulatory models. Our findings indicate that the contextual embeddings learned by LLMs can serve as an explicit numerical model of the shared, context-rich meaning space humans use to communicate their thoughts to one another.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Generalized convulsive seizures (GCSs) are the main risk factor of sudden unexpected death in epilepsy (SUDEP), which is likely due to peri-ictal cardiorespiratory dysfunction. The incidence of GCS-induced cardiac arrhythmias, their relationship to seizure severity markers, and their role in SUDEP physiopathology are unknown. The aim of this study was to analyze the incidence of seizure-induced cardiac arrhythmias, their association with electroclinical features and seizure severity biomarkers, as well as their specific occurrences in SUDEP cases. METHODS:-score test for 2 population proportions was used to test whether the proportion of seizures and patients with postconvulsive ESAWB or bradycardia differed between SUDEP cases and survivors. RESULTS:> 0.05). DISCUSSION/CONCLUSIONS:Markers of seizure severity are not related to seizure-induced arrhythmias of interest, suggesting that other factors such as occult cardiac abnormalities may be relevant for their occurrence. Seizure-induced ESAWB and bradycardia were more frequent in SUDEP cases, although this observation was based on a very limited number of SUDEP patients. Further case-control studies are needed to evaluate the yield of arrhythmias of interest along with respiratory changes as potential SUDEP biomarkers.

Epilepsy is often comorbid with psychiatric illnesses, including anxiety and depression. Despite the high incidence of psychiatric comorbidities in people with epilepsy, few studies address the underlying mechanisms. Stress can trigger epilepsy and depression. Evidence from human and animal studies supports that hypothalamic-pituitary-adrenal (HPA) axis dysfunction may contribute to both disorders and their comorbidity ( Kanner, 2003). Here, we investigate if HPA axis dysfunction may influence epilepsy outcomes and psychiatric comorbidities. We generated a novel mouse model (Kcc2/Crh KO mice) lacking the K⁺/Cl^- cotransporter, KCC2, in corticotropin-releasing hormone (CRH) neurons, which exhibit stress- and seizure-induced HPA axis hyperactivation ( Melon et al., 2018). We used the Kcc2/Crh KO mice to examine the impact on epilepsy outcomes, including seizure frequency/burden, comorbid behavioral deficits, and sudden unexpected death in epilepsy (SUDEP) risk. We found sex differences in HPA axis dysfunction's effect on chronically epileptic KCC2/Crh KO mice seizure burden, vulnerability to comorbid behavioral deficits, and SUDEP. Suppressing HPA axis hyperexcitability in this model using pharmacological or chemogenetic approaches decreased SUDEP incidence, suggesting that HPA axis dysfunction may contribute to SUDEP. Altered neuroendocrine markers were present in SUDEP cases compared with people with epilepsy or individuals without epilepsy. Together, these findings implicate HPA axis dysfunction in the pathophysiological mechanisms contributing to psychiatric comorbidities in epilepsy and SUDEP.

Accurate seizure and epilepsy diagnosis remains a challenging task due to the complexity and variability of manifestations, which can lead to delayed or missed diagnosis. Machine learning (ML) and artificial intelligence (AI) is a rapidly developing field, with growing interest in integrating and applying these tools to aid clinicians facing diagnostic uncertainties. ML algorithms, particularly deep neural networks, are increasingly employed in interpreting electroencephalograms (EEG), neuroimaging, wearable data, and seizure videos. This review discusses the development and testing phases of AI/ML tools, emphasizing the importance of generalizability and interpretability in medical applications, and highlights recent publications that demonstrate the current and potential utility of AI to aid clinicians in diagnosing epilepsy. Current barriers of AI integration in patient care include dataset availability and heterogeneity, which limit studies' quality, interpretability, comparability, and generalizability. ML and AI offer substantial promise in improving the accuracy and efficiency of epilepsy diagnosis. The growing availability of diverse datasets, enhanced processing speed, and ongoing efforts to standardize reporting contribute to the evolving landscape of AI applications in clinical care.

Studies on epilepsy mortality in the United States are limited. We used the National Vital Statistics System Multiple Cause of Death data to investigate mortality rates and trends during 2011-2021 for epilepsy (defined by the International Classification of Diseases, 10th Revision, codes G40.0-G40.9) as an underlying, contributing, or any cause of death (i.e., either an underlying or contributing cause) for U.S. residents. We also examined epilepsy as an underlying or contributing cause of death by selected sociodemographic characteristics to assess mortality rate changes and disparities in subpopulations. During 2011-2021, the overall age-standardized mortality rates for epilepsy as an underlying (39 % of all deaths) or contributing (61 % of all deaths) cause of death increased 83.6 % (from 2.9 per million to 6.4 per million population) as underlying cause and 144.1 % (from 3.3 per million to 11.0 per million population) as contributing cause (P < 0.001 for both based on annual percent changes). Compared to 2011-2015, in 2016-2020 mortality rates with epilepsy as an underlying or contributing cause of death were higher overall and in nearly all subgroups. Overall, mortality rates with epilepsy as an underlying or contributing cause of death were higher in older age groups, among males than females, among non-Hispanic Black or non-Hispanic American Indian/Alaska Native persons than non-Hispanic White persons, among those living in the West and Midwest than those living in the Northeast, and in nonmetro counties compared to urban regions. Results identify priority subgroups for intervention to reduce mortality in people with epilepsy and eliminate mortality disparity.

Neural responses in visual cortex adapt to prolonged and repeated stimuli. While adaptation occurs across the visual cortex, it is unclear how adaptation patterns and computational mechanisms differ across the visual hierarchy. Here we characterize two signatures of short-term neural adaptation in time-varying intracranial electroencephalography (iEEG) data collected while participants viewed naturalistic image categories varying in duration and repetition interval. Ventral- and lateral-occipitotemporal cortex exhibit slower and prolonged adaptation to single stimuli and slower recovery from adaptation to repeated stimuli compared to V1-V3. For category-selective electrodes, recovery from adaptation is slower for preferred than non-preferred stimuli. To model neural adaptation we augment our delayed divisive normalization (DN) model by scaling the input strength as a function of stimulus category, enabling the model to accurately predict neural responses across multiple image categories. The model fits suggest that differences in adaptation patterns arise from slower normalization dynamics in higher visual areas interacting with differences in input strength resulting from category selectivity. Our results reveal systematic differences in temporal adaptation of neural population responses between lower and higher visual brain areas and show that a single computational model of history-dependent normalization dynamics, fit with area-specific parameters, accounts for these differences.

Although sudden unexpected death in epilepsy (SUDEP) is the most feared epilepsy outcome, there is a dearth of SUDEP counseling provided by neurologists. This may reflect limited time, as well as the lack of guidance on the timing and structure for counseling. We evaluated records from SUDEP cases to examine frequency of inpatient and outpatient SUDEP counseling, and whether counseling practices were influenced by risk factors and biomarkers, such as post-ictal generalized EEG suppression (PGES). We found a striking lack of SUDEP counseling despite modifiable SUDEP risk factors; counseling was limited to outpatients despite many patients having inpatient visits within a year of SUDEP. PGES was inconsistently documented and was never included in counseling. There is an opportunity to greatly improve SUDEP counseling by utilizing inpatient settings and prompting algorithms incorporating risk factors and biomarkers.

Racial disparities affect multiple dimensions of epilepsy care including epilepsy surgery. This study aims to further explore these disparities by determining the utilization of invasive neuromodulation devices according to race and ethnicity in a multicenter study of patients living with focal drug-resistant epilepsy (DRE). We performed a post hoc analysis of the Human Epilepsy Project 2 (HEP2) data. HEP2 is a prospective study of patients living with focal DRE involving 10 sites distributed across the United States. There were no statistical differences in the racial distribution of the study population compared to the US population using census data except for patients reporting more than one race. Of 154 patients enrolled in HEP2, 55 (36%) underwent invasive neuromodulation for DRE management at some point in the course of their epilepsy. Of those, 36 (71%) were patients who identified as White. Patients were significantly less likely to have a device if they identified solely as Black/African American than if they did not (odds ratio = .21, 95% confidence interval = .05-.96, p = .03). Invasive neuromodulation for management of DRE is underutilized in the Black/African American population, indicating a new facet of racial disparities in epilepsy care.

Contextual embeddings, derived from deep language models (DLMs), provide a continuous vectorial representation of language. This embedding space differs fundamentally from the symbolic representations posited by traditional psycholinguistics. We hypothesize that language areas in the human brain, similar to DLMs, rely on a continuous embedding space to represent language. To test this hypothesis, we densely record the neural activity patterns in the inferior frontal gyrus (IFG) of three participants using dense intracranial arrays while they listened to a 30-minute podcast. From these fine-grained spatiotemporal neural recordings, we derive a continuous vectorial representation for each word (i.e., a brain embedding) in each patient. Using stringent zero-shot mapping we demonstrate that brain embeddings in the IFG and the DLM contextual embedding space have common geometric patterns. The common geometric patterns allow us to predict the brain embedding in IFG of a given left-out word based solely on its geometrical relationship to other non-overlapping words in the podcast. Furthermore, we show that contextual embeddings capture the geometry of IFG embeddings better than static word embeddings. The continuous brain embedding space exposes a vector-based neural code for natural language processing in the human brain.