Prospective Validation of Neurophysiologic Outcome Prediction in Acute Brain Injury

急性脑损伤神经生理结果预测的前瞻性验证

• 批准号: 10584338
• 负责人: Aaron F Struck
• 金额: $ 81.38万
• 依托单位: BETH ISRAEL DEACONESS MEDICAL CENTER
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-01-01 至 2027-12-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10584338
• 关键词: Acute; Acute Brain Injuries; Address; Anticonvulsants; Artificial Intelligence; Binding; Bioinformatics; Biological Markers; Brain; Brain Injuries; Caring; Cerebrum; Cessation of life; Chronic; Chronic Phase; Clinical; Clinical Data; Clinical Trials; Collaborations; Computerized Medical Record; Critical Illness; Data; Data Collection; Databases; Development; Electroencephalography; Epilepsy; Epileptogenesis; Equipoise; Future; General Hospitals; Geography; Guidelines; Hospitalization; Hospitals; Image; Injury; Learning; Machine Learning; Massachusetts; Measures; Modeling; Monitor; Neurological outcome; Observational Study; Outcome; Pathologic; Patient-Focused Outcomes; Patients; Pattern; Periodicals; Pharmaceutical Preparations; Phase; Process; Prognosis; Property; Psychological reinforcement; Resources; Retrospective Studies; Risk; Risk Factors; Sample Size; Secondary Prevention; Seizures; Site; Techniques; Time; Translating; Uncertainty; Universities; Update; Validation; Wisconsin; Workload; acquired epilepsy; bioinformatics infrastructure; brain electrical activity; care outcomes; clinical care; clinically actionable; clinically significant; cohort; deep learning model; detector; disability; epileptiform; functional improvement; improved; improved outcome; large scale data; machine learning algorithm; machine learning model; multimodality; neurological recovery; neurophysiology; novel; outcome prediction; predictive modeling; prevent; prospective; public database; radiological imaging; risk prediction; signal processing; tool

PROJECT SUMMARY Acute brain injury causes over 163,000 deaths in the US annually and leaves many more patients with long- term disability. Preventing secondary brain injury is critical to improving neurologic outcomes in these patients. Pathological brain electrical activity (measured via EEG) following acute brain injury contributes to long-term disability via seizures in the acute phase and epilepsy in the chronic phase. EEG is the primary tool for monitoring aberrant brain activity, yet it is underutilized due to uncertainty regarding the clinical significance of EEG findings and the high workload associated with interpreting high volumes of EEG data. Our group has made progress toward addressing these short-comings by developing three novel machine learning algorithms: 1) a seizure forecasting model for hospitalized patients (“2HELPS2B”); 2) a model that measures the Burden of Epileptiform Activity (EA—seizures and highly epileptiform patterns such as lateralized periodic discharges) to predict neurologic outcomes (“BEACON”); and 3) A model that uses EA in the acute phase of brain injury to predict the risk of developing epilepsy. This proposal is the culminating step required to translate preliminary studies into actionable clinical tools. In this prospective multicenter observational study, we will collect clinical and EEG data on 3000 patients with acute brain injury to further develop and validate these models. Our study leverages an existing multicenter bioinformatics infrastructure and established collaborations between Yale University, University of Wisconsin, and Massachusetts General Hospital, to create the scale and quality of data needed to address three specific aims: SA1: Develop and prospectively validate an automated in-hospital seizure-forecasting model for use in acute brain injury, based on our previously developed 2HELPS2B score—termed auto-2HELPS2B; SA2: Prospectively validate the BEACON model for the impact of prolonged epileptiform activity on functional and clinical outcomes in critical illness at discharge, 3 months, 6 months, 1-year, and 2-years, and estimate the optimal anti-seizure drug administration strategy (indications and intensity of drugs) to mitigate detrimental effects of EA on functional and clinical outcomes; SA3: Prospectively validate EAs as a biomarker of 1- and 2- year epilepsy risk after acute brain injury, evaluate effects of anti-seizure drugs on acute phase EAs and subsequent development of epilepsy, and combine EEG with radiographic and clinical information to further improve our current epilepsy forecasting risk model. Upon completing these aims, we will have 1. A large and representative database of high-quality EEG and clinical data on brain-injured patients with 2-year of outcomes data. 2. Validated EEG tools to guide care in determining acute seizure risk, prognosis for neurological recovery, and the likelihood of developing epilepsy.

项目概要 在美国，急性脑损伤每年导致超过 163,000 人死亡，并导致更多患者长期患病。 预防继发性脑损伤对于改善这些患者的神经系统预后至关重要。 急性脑损伤后的病理性脑电活动（通过脑电图测量）有助于长期 急性期癫痫发作和慢性期癫痫是诊断残疾的主要工具。 监测异常的大脑活动，但由于临床意义的不确定性，它未被充分利用 脑电图结果以及与解释大量脑电图数据相关的高工作量。 我们的团队通过开发三款新型机器，在解决这些缺点方面取得了进展 学习算法：1）住院患者癫痫发作预测模型（“2HELPS2B”）；2） 测量癫痫样活动的负担（EA — 癫痫发作和高度癫痫样模式，例如偏侧化 周期性放电）来预测神经系统结果（“BEACON”）；以及 3) 在急性发作中使用 EA 的模型； 脑损伤阶段可预测患癫痫的风险。 该提案是将初步研究转化为可行的临床工具所需的最终步骤。 这项前瞻性多中心观察研究，我们将收集 3000 名患有以下疾病的患者的临床和脑电图数据： 我们的研究利用现有的多中心急性脑损伤来进一步开发和验证这些模型。 生物信息学基础设施以及耶鲁大学、威斯康星大学、 和马萨诸塞州总医院，创建解决三个特定问题所需的规模和质量的数据 目标：SA1：开发并前瞻性验证自动院内癫痫发作预测模型，用于 急性脑损伤，基于我们之前开发的 2HELPS2B 评分（称为 auto-2HELPS2B）； 前瞻性验证 BEACON 模型对于长期癫痫样活动对功能和功能的影响 出院时、3个月、6个月、1年和2年的危重症临床结果，并估计 减轻疼痛的最佳抗癫痫药物给药策略（药物的适应症和强度） EA 对功能和临床结果的影响；SA3：前瞻性验证 EA 作为 1- 和 2- 的生物标志物 急性脑损伤后的癫痫风险，评估抗癫痫药物对急性期 EA 的影响和 癫痫的后续发展，并将脑电图与放射线和临床信息相结合，以进一步 改进我们当前的癫痫预测风险模型。 完成这些目标后，我们将拥有 1. 一个大型且具有代表性的高质量脑电图和 脑损伤患者的临床数据以及 2 年结果数据 2. 经验证的脑电图工具可指导护理。 确定急性癫痫发作风险、神经恢复预后以及发生癫痫的可能性。