Artificial intelligence (AI) has demonstrated significant potential in electrocardiogram (ECG) analysis and cardiovascular disease assessment. Recently, foundation models have played a remarkable role in advancing medical AI, bringing benefits such as efficient disease diagnosis and cross-domain knowledge transfer. The development of an ECG foundation model holds the promise of elevating AI-ECG research to new heights. However, building such a model poses several challenges, including insufficient database sample sizes and inadequate generalization across multiple domains. In addition, there is a notable performance gap between single-lead and multilead ECG analysis. We propose a general-purpose ECG foundation model (ECGFounder), which leverages real-world ECG annotations from cardiologists to broaden the diagnostic capabilities of ECG analysis. ECGFounder was built on 10,771,552 ECGs from 1,818,247 unique subjects with 150 label categories from the Harvard-Emory ECG Database, enabling comprehensive cardiovascular disease diagnosis. The model is designed to be both an effective out-of-the-box solution and easily fine-tunable for downstream tasks, maximizing usability. Importantly, we extended its application to reduced-lead ECGs, particularly single-lead ECGs. ECGFounder is therefore applicable to various downstream tasks in mobile and remote monitoring scenarios. Experimental results demonstrate that ECGFounder achieves expert-level performance on internal validation sets, with area under the receiver operating characteristic curve (AUROC) exceeding 0.95 for 80 diagnoses. It also shows strong classification performance and generalization across various diagnoses on external validation sets. When fine-tuned, ECGFounder outperforms baseline models in demographic analysis, clinical event detection, and cross-modality cardiac rhythm diagnosis, surpassing baseline methods by 3 to 5 points in the AUROC. The ECG foundation model offers an effective solution, allowing it to generalize across a wide range of tasks. By enhancing existing cardiovascular diagnostics and facilitating integration with cloud-based systems, which analyze ECG data uploaded from wearable devices, it significantly contributes to the advancement of the cardiovascular AI community and enables management of cardiac conditions. (Funded by the National Science Foundation and others.).

Background and Objectives Delayed cerebral ischemia (DCI) is the leading complication of subarachnoid hemorrhage (SAH). Because DCI was traditionally thought to be caused by large vessel vasospasm, transcranial Doppler ultrasounds (TCDs) have been the standard of care. Continuous EEG has emerged as a promising complementary monitoring modality and predicts increased DCI risk. Our objective was to determine whether combining EEG and TCD data improves prediction of DCI after SAH. We hypothesize that integrating these diagnostic modalities improves DCI prediction. Methods We retrospectively assessed patients with moderate to severe SAH (2011–2015; Fisher 3–4 or Hunt-Hess 4–5) who had both prospective TCD and EEG acquisition during hospitalization. Middle cerebral artery (MCA) peak systolic velocities (PSVs) and the presence or absence of epileptiform abnormalities (EAs), defined as seizures, epileptiform discharges, and rhythmic/periodic activity, were recorded daily. Logistic regressions were used to identify significant covariates of EAs and TCD to predict DCI. Group-based trajectory modeling (GBTM) was used to account for changes over time by identifying distinct group trajectories of MCA PSV and EAs associated with DCI risk. Results We assessed 107 patients; DCI developed in 56 (51.9%). Univariate predictors of DCI are presence of high-MCA velocity (PSV ≥200 cm/s, sensitivity 27%, specificity 89%) and EAs (sensitivity 66%, specificity 62%) on or before day 3. Two univariate GBTM trajectories of EAs predicted DCI (sensitivity 64%, specificity 62.75%). Logistic regression and GBTM models using both TCD and EEG monitoring performed better. The best logistic regression and GBTM models used both TCD and EEG data, Hunt-Hess score at admission, and aneurysm treatment as predictors of DCI (logistic regression: sensitivity 90%, specificity 70%; GBTM: sensitivity 89%, specificity 67%). Discussion EEG and TCD biomarkers combined provide the best prediction of DCI. The conjunction of clinical variables with the timing of EAs and high MCA velocities improved model performance. These results suggest that TCD and cEEG are promising complementary monitoring modalities for DCI prediction. Our model has potential to serve as a decision support tool in SAH management. Classification of Evidence This study provides Class II evidence that combined TCD and EEG monitoring can identify delayed cerebral ischemia after SAH.

Objective: To determine whether patients could self-report physical and mental health assessments in the waiting room and whether these assessments would be associated with modified Rankin Scale (mRS) and Quality of Life in Epilepsy (QOLIE-10) scores. Methods: We offered iPad-based surveys to consecutive adult neurology patients at check-in to collect patient-reported outcome measures (PROMs). We collected demographic and clinical data on 6,075 patients through survey or administrative claims and PROMs from participating patients. We compared demographic characteristics of participants and nonparticipants and tested associations between physical and mental health scores and mRS and QOLIE-10. Results: Of 6,075 patients seen by neurologists during the study period, 2,992 (49.3%) participated in the survey. Compared to nonparticipating patients, participating patients more often were privately insured (53.5% vs 42.7%, p < 0.01), married (51.5% vs 47.9%, p < 0.01), and seen in general neurology (nonsubspecialty) clinics (53.1% vs 46.6%, p < 0.01) and more likely to report English as their preferred language (50.1% vs 38.4%, p < 0.01). Participating patients had a mean physical health T score of 28.7 (SD 15) and mental health T score of 33 (SD 15), which were 3 and 2 SD worse than the average for the US general population, respectively. Mean T scores in every category of the mRS were different from every other category (n = 232, p < 0.01). Patient Reported Outcomes Measurement Information System-10 T scores were linearly associated with QOLIE-10 scores (n = 202, p < 0.01) Conclusions: Systematic digital collection of PROMs is feasible. Differences among survey participants and nonparticipants highlight the need to develop multilingual measurement tools that may improve collection from vulnerable populations.

Summary: The purpose of this study is to evaluate automated implementations of continuous EEG monitoring-based detection of delayed cerebral ischemia based on methods used in classical retrospective studies. We studied 95 patients with either Fisher 3 or Hunt Hess 4 to 5 aneurysmal subarachnoid hemorrhage who were admitted to the Neurosciences ICU and underwent continuous EEG monitoring. We implemented several variations of two classical algorithms for automated detection of delayed cerebral ischemia based on decreases in alpha-delta ratio and relative alpha variability. Of 95 patients, 43 (45%) developed delayed cerebral ischemia. Our automated implementation of the classical alpha-delta ratio-based trending method resulted in a sensitivity and specificity (Se,Sp) of (80,27)%, compared with the values of (100,76)% reported in the classic study using similar methods in a nonautomated fashion. Our automated implementation of the classical relative alpha variability-based trending method yielded (Se,Sp) values of (65,43)%, compared with (100,46)% reported in the classic study using nonautomated analysis. Our findings suggest that improved methods to detect decreases in alpha-delta ratio and relative alpha variability are needed before an automated EEG-based early delayed cerebral ischemia detection system is ready for clinical use.

To characterize the amount of EEG suppression achieved in refractory status epilepticus (RSE) patients treated with pharmacologically-induced coma (PIC). We analyzed EEG recordings from 35 RSE patients between 21–84 years-old who received PIC that target burst suppression and quantified the amount of EEG suppression using the burst suppression probability (BSP). Then we measured the variability of BSPs with respect to a reference level of BSP 0.8 ± 0.15. Finally, we also measured the variability of BSPs with respect to the amount of intravenous anesthetic drugs (IVADs) received by the patients. Patients remained in the reference BSP range for only 8% (median, interquartile range IQR [0, 29] %) of the total time under treatment. The median time with BSP below the reference range was 84% (IQR [37, 100] %). BSPs in some patients drifted significantly over time despite constant infusion rates of IVADs. Similar weight-normalized infusion rates of IVADs in different patients nearly always resulted in distinct BSPs (probability 0.93 (IQR [0.82, 1.0]). This study quantitatively identified high variability in the amount of EEG suppression achieved in clinical practice when treating RSE patients. While some of this variability may arise from clinicians purposefully deviating from clinical practice guidelines, our results show that the high variability also arises in part from significant inter- and intra- individual pharmacokinetic/pharmacodynamic variation. Our results indicate that the delicate balance between maintaining sufficient EEG suppression in RSE patients and minimizing IVAD exposure in clinical practice is challenging to achieve. This may affect patient outcomes and confound studies seeking to determine an optimal amount of EEG suppression for treatment of RSE. Therefore, our analysis points to the need for developing an alternative paradigm, such as vigilant anesthetic management as happens in operating rooms, or closed-loop anesthesia delivery, for investigating and providing induced-coma therapy to RSE patients.