Extracting seizure frequency from epilepsy clinic notes: a machine reading approach to natural language processing.

OBJECTIVE: Seizure frequency and seizure freedom are among the most important outcome measures for patients with epilepsy. In this study, we aimed to automatically extract this clinical information from unstructured text in clinical notes. If successful, this could improve clinical decision-making in epilepsy patients and allow for rapid, large-scale retrospective research. MATERIALS AND METHODS: We developed a finetuning pipeline for pretrained neural models to classify patients as being seizure-free and to extract text containing their seizure frequency and date of last seizure from clinical notes. We annotated 1000 notes for use as training and testing data and determined how well 3 pretrained neural models, BERT, RoBERTa, and Bio_ClinicalBERT, could identify and extract the desired information after finetuning. RESULTS: The finetuned models (BERTFT, Bio_ClinicalBERTFT, and RoBERTaFT) achieved near-human performance when classifying patients as seizure free, with BERTFT and Bio_ClinicalBERTFT achieving accuracy scores over 80%. All 3 models also achieved human performance when extracting seizure frequency and date of last seizure, with overall F1 scores over 0.80. The best combination of models was Bio_ClinicalBERTFT for classification, and RoBERTaFT for text extraction. Most of the gains in performance due to finetuning required roughly 70 annotated notes. DISCUSSION AND CONCLUSION: Our novel machine reading approach to extracting important clinical outcomes performed at or near human performance on several tasks. This approach opens new possibilities to support clinical practice and conduct large-scale retrospective clinical research. Future studies can use our finetuning pipeline with minimal training annotations to answer new clinical questions.

Simulated diagnostic performance of low-field MRI: Harnessing open-access datasets to evaluate novel devices.

The purpose of this study is to demonstrate a method for virtually evaluating novel imaging devices using machine learning and open-access datasets, here applied to a new, low-field strength portable 64mT MRI device. Paired 3 T and 64mT brain images were used to develop and validate a transformation converting standard clinical images to low-field quality images. Separately, 3 T images were aggregated from open-source databases spanning four neuropathologies: low-grade glioma (LGG, N = 76), high-grade glioma (HGG, N = 259), stroke (N = 28), and multiple sclerosis (MS, N = 20). The transformation method was then applied to the open-source data to generate simulated low-field images for each pathology. Convolutional neural networks (DenseNet-121) were trained to detect pathology in axial slices from either 3 T or simulated 64 mT images, and their relative performance was compared to characterize the potential diagnostic capabilities of low-field imaging. Algorithm performance was measured using area under the receiver operating characteristic curve. Across all cohorts, pathology detection was similar between 3 T and simulated 64mT images (LGG: 0.97 vs. 0.98; HGG: 0.96 vs. 0.95; stroke: 0.94 vs. 0.94; MS: 0.90 vs 0.87). Pathology detection was further characterized as a function of lesion size, intensity, and contrast. Simulated images showed decreasing sensitivity for lesions smaller than 4 cm2. While simulations cannot replace prospective trials during the evaluation of medical devices, they can provide guidance and justification for prospective studies. Simulated data derived from open-source imaging databases may facilitate testing and validation of new imaging devices.

A Full-Stack Application for Detecting Seizures and Reducing Data During Continuous Electroencephalogram Monitoring.

Continuous electroencephalogram monitoring is associated with lower mortality in critically ill patients; however, it is underused due to the resource-intensive nature of manually interpreting prolonged streams of continuous electroencephalogram data. Here, we present a novel real-time, machine learning-based alerting and monitoring system for epilepsy and seizures that dramatically reduces the amount of manual electroencephalogram review. METHODS: We developed a custom data reduction algorithm using a random forest and deployed it within an online cloud-based platform, which streams data and communicates interactively with caregivers via a web interface to display algorithm results. We developed real-time, machine learning-based alerting and monitoring system for epilepsy and seizures on continuous electroencephalogram recordings from 77 patients undergoing routine scalp ICU electroencephalogram monitoring and tested it on an additional 20 patients. RESULTS: We achieved a mean seizure sensitivity of 84% in cross-validation and 85% in testing, as well as a mean specificity of 83% in cross-validation and 86% in testing, corresponding to a high level of data reduction. This study validates a platform for machine learning-assisted continuous electroencephalogram analysis and represents a meaningful step toward improving utility and decreasing cost of continuous electroencephalogram monitoring. We also make our high-quality annotated dataset of 97 ICU continuous electroencephalogram recordings public for others to validate and improve upon our methods.

Can antiepileptic efficacy and epilepsy variables be studied from electronic health records? A review of current approaches.

As automated data extraction and natural language processing (NLP) are rapidly evolving, improving healthcare delivery by harnessing large data is garnering great interest. Assessing antiepileptic drug (AED) efficacy and other epilepsy variables pertinent to healthcare delivery remain a critical barrier to improving patient care. In this systematic review, we examined automatic electronic health record (EHR) extraction methodologies pertinent to epilepsy. We also reviewed more generalizable NLP pipelines to extract other critical patient variables. Our review found varying reports of performance measures. Whereas automated data extraction pipelines are a crucial advancement, this review calls attention to standardizing NLP methodology and accuracy reporting for greater generalizability. Moreover, the use of crowdsourcing competitions to spur innovative NLP pipelines would further advance this field.

Model-based design for seizure control by stimulation.

OBJECTIVE: Current brain stimulation paradigms are largely empirical rather than theoretical. An opportunity exists to improve upon their modest effectiveness in closed-loop control strategies with the development of theoretically grounded, model-based designs. APPROACH: Inspired by this need, here we couple experimental data and mathematical modeling with a control-theoretic strategy for seizure termination. We begin by exercising a dynamical systems approach to model seizures (n = 94) recorded using intracranial EEG (iEEG) from 21 patients with medication-resistant, localization-related epilepsy. MAIN RESULTS: Although each patient's seizures displayed unique spatial and temporal patterns, their evolution can be parsimoniously characterized by the same model form. Idiosyncracies of the model can inform individualized intervention strategies, specifically in iEEG samples with well-localized seizure onset zones. Temporal fluctuations in the spatial profiles of the oscillatory modes show that seizure onset marks a transition into a regime in which the underlying system supports prolonged rhythmic and focal activity. Based on these observations, we propose a control-theoretic strategy that aims to stabilize ictal activity using static output feedback for linear time-invariant switching systems. Finally, we demonstrate in silico that our proposed strategy allows us to dampen the emerging focal oscillatory sources using only a small set of electrodes. SIGNIFICANCE: Our integrative study informs the development of modulation and control algorithms for neurostimulation that could improve the effectiveness of implantable, closed-loop anti-epileptic devices.

IRIS: A Modular Platform for Continuous Monitoring and Caretaker Notification in the Intensive Care Unit.

OBJECTIVE: New approaches are needed to interpret large amounts of physiologic data continuously recorded in the ICU. We developed and prospectively validated a versatile platform (IRIS) for real-time ICU physiologic monitoring, clinical decision making, and caretaker notification. METHODS: IRIS was implemented in the neurointensive care unit to stream multimodal time series data, including EEG, intracranial pressure (ICP), and brain tissue oxygenation (PbtO2), from ICU monitors to an analysis server. IRIS was applied for 364 patients undergoing continuous EEG, 26 patients undergoing burst suppression monitoring, and four patients undergoing intracranial pressure and brain tissue oxygen monitoring. Custom algorithms were used to identify periods of elevated ICP, compute burst suppression ratios (BSRs), and detect faulty or disconnected EEG electrodes. Hospital staff were notified of clinically relevant events using our secure API to route alerts through a password-protected smartphone application. RESULTS: Sustained increases in ICP and concordant decreases in PbtO2 were reliably detected using user-defined thresholds and alert throttling. BSR trends computed by the platform correlated highly with manual neurologist markings (r2 0.633-0.781; p < 0.0001). The platform identified EEG electrodes with poor signal quality with 95% positive predictive value, and reduced latency of technician response by 93%. CONCLUSION: This study validates a flexible real-time platform for monitoring and interpreting ICU data and notifying caretakers of actionable results, with potential to reduce the manual burden of continuous monitoring services on care providers. SIGNIFICANCE: This work represents an important step toward facilitating translational medical data analytics to improve patient care and reduce health care costs.

Big data in status epilepticus.

Status epilepticus care and treatment are already being touched by the revolution in data science. New approaches designed to leverage the tremendous potential of "big data" in the clinical sphere are enabling researchers and clinicians to extract information from sources such as administrative claims data, the electronic medical health record, and continuous physiologic monitoring data streams. Algorithmic methods of data extraction also offer potential to fuse multimodal data (including text-based documentation, imaging data, and time-series data) to improve patient assessment and stratification beyond the manual capabilities of individual physicians. Still, the potential of data science to impact the diagnosis, treatment, and minute-to-minute care of patients with status epilepticus is only starting to be appreciated. In this brief review, we discuss how data science is impacting the field and draw examples from the following three main areas: (1) analysis of insurance claims from large administrative datasets to evaluate the impact of continuous electroencephalogram (EEG) monitoring on clinical outcomes; (2) natural language processing of the electronic health record to find, classify, and stratify patients for prognostication and treatment; and (3) real-time systems for data analysis, data reduction, and multimodal data fusion to guide therapy in real time. While early, it is our hope that these examples will stimulate investigators to leverage data science, computer science, and engineering methods to improve the care and outcome of patients with status epilepticus and other neurological disorders. This article is part of the Special Issue "Proceedings of the 7th London-Innsbruck Colloquium on Status Epilepticus and Acute Seizures".

Machine learning to detect signatures of disease in liquid biopsies - a user's guide.

New technologies that measure sparse molecular biomarkers from easily accessible bodily fluids (e.g. blood, urine, and saliva) are revolutionizing disease diagnostics and precision medicine. Microchip devices can measure more disease biomarkers with better sensitivity and specificity each year, but clinical interpretation of these biomarkers remains a challenge. Single biomarkers in 'liquid biopsy' often cannot accurately predict the state of a disease due to heterogeneity in phenotype and disease expression across individuals. To address this challenge, investigators are combining multiplexed measurements of different biomarkers that together define robust signatures for specific disease states. Machine learning is a useful tool to automatically discover and detect these signatures, especially as new technologies output increasing quantities of molecular data. In this paper, we review the state of the field of machine learning applied to molecular diagnostics and provide practical guidance to use this tool effectively and to avoid common pitfalls.

Intracranial EEG fluctuates over months after implanting electrodes in human brain.

OBJECTIVE: Implanting subdural and penetrating electrodes in the brain causes acute trauma and inflammation that affect intracranial electroencephalographic (iEEG) recordings. This behavior and its potential impact on clinical decision-making and algorithms for implanted devices have not been assessed in detail. In this study we aim to characterize the temporal and spatial variability of continuous, prolonged human iEEG recordings. APPROACH: Intracranial electroencephalography from 15 patients with drug-refractory epilepsy, each implanted with 16 subdural electrodes and continuously monitored for an average of 18 months, was included in this study. Time and spectral domain features were computed each day for each channel for the duration of each patient's recording. Metrics to capture post-implantation feature changes and inflexion points were computed on group and individual levels. A linear mixed model was used to characterize transient group-level changes in feature values post-implantation and independent linear models were used to describe individual variability. MAIN RESULTS: A significant decline in features important to seizure detection and prediction algorithms (mean line length, energy, and half-wave), as well as mean power in the Berger and high gamma bands, was observed in many patients over 100 d following implantation. In addition, spatial variability across electrodes declines post-implantation following a similar timeframe. All selected features decreased by 14-50% in the initial 75 d of recording on the group level, and at least one feature demonstrated this pattern in 13 of the 15 patients. Our findings indicate that iEEG signal features demonstrate increased variability following implantation, most notably in the weeks immediately post-implant. SIGNIFICANCE: These findings suggest that conclusions drawn from iEEG, both clinically and for research, should account for spatiotemporal signal variability and that properly assessing the iEEG in patients, depending upon the application, may require extended monitoring.

Crowdsourcing seizure detection: algorithm development and validation on human implanted device recordings.

There exist significant clinical and basic research needs for accurate, automated seizure detection algorithms. These algorithms have translational potential in responsive neurostimulation devices and in automatic parsing of continuous intracranial electroencephalography data. An important barrier to developing accurate, validated algorithms for seizure detection is limited access to high-quality, expertly annotated seizure data from prolonged recordings. To overcome this, we hosted a kaggle.com competition to crowdsource the development of seizure detection algorithms using intracranial electroencephalography from canines and humans with epilepsy. The top three performing algorithms from the contest were then validated on out-of-sample patient data including standard clinical data and continuous ambulatory human data obtained over several years using the implantable NeuroVista seizure advisory system. Two hundred teams of data scientists from all over the world participated in the kaggle.com competition. The top performing teams submitted highly accurate algorithms with consistent performance in the out-of-sample validation study. The performance of these seizure detection algorithms, achieved using freely available code and data, sets a new reproducible benchmark for personalized seizure detection. We have also shared a 'plug and play' pipeline to allow other researchers to easily use these algorithms on their own datasets. The success of this competition demonstrates how sharing code and high quality data results in the creation of powerful translational tools with significant potential to impact patient care.

Topological distortion and reorganized modular structure of gut microbial co-occurrence networks in inflammatory bowel disease.

The gut microbiome plays a key role in human health, and alterations of the normal gut flora are associated with a variety of distinct disease states. Yet, the natural dependencies between microbes in healthy and diseased individuals remain far from understood. Here we use a network-based approach to characterize microbial co-occurrence in individuals with inflammatory bowel disease (IBD) and healthy (non-IBD control) individuals. We find that microbial networks in patients with IBD differ in both global structure and local connectivity patterns. While a "core" microbiome is preserved, network topology of other densely interconnected microbe modules is distorted, with potent inflammation-mediating organisms assuming roles as integrative and highly connected inter-modular hubs. We show that while both networks display a rich-club organization, in which a small set of microbes commonly co-occur, the healthy network is more easily disrupted by elimination of a small number of key species. Further investigation of network alterations in disease might offer mechanistic insights into the specific pathogens responsible for microbiome-mediated inflammation in IBD.