Determining the clinical utility of BioEP: a multisite, prospective, study

我们的建议侧重于癫痫：一种严重的神经系统疾病。癫痫的特点是有反复、自发发作的倾向。由于癫痫发作似乎是随机发生的，癫痫的诊断是非常有挑战性的。脑电图（EEG）是一个重要的临床工具，揭示异常，表明癫痫。然而，约70%记录的脑电图没有显示癫痫的临床指征。因此，诊断延误通常超过一年，误诊率估计为30%。据估计，这种误诊每年给NHS造成的损失超过3亿英镑。为了应对这一挑战，神经系统学开发了一种癫痫的计算机生物标志物：BioEP。BioEP的工作原理是将常规采集的脑电图整合到大脑的计算机模型中。对该模型的模拟，使用先进的机器学习算法进行查询，提供了一个风险评分，显示癫痫发作的容易程度。对于目前临床认为无贡献的脑电图，BioEP风险评分提供了有价值的附加诊断信息。支持BioEP的研究得到了《柳叶刀神经病学》（2015年作为癫痫领域的突破性研究）、癫痫基金会（2017年获得商业化奖）和《自然》（神经学是2021年衍生奖的最后四名之一）的认可。通过Innovate UK和NIHR的资助，Neuronostics开发了一种由BioEP驱动的诊断决策支持工具的原型，可以实现半自动特征提取和报告撰写。神经系统学最近完成了一项回顾性研究（使用来自已知和准确诊断的患者的数据），以验证BioEP的性能。这是同类研究中规模最大的，我们证明了BioEP可以揭示明显正常的脑电图对癫痫的易感性，其水平超过了目前的临床实践。在目前的项目中，我们将进行一项前瞻性研究，以证明BioEP风险评分的临床实用性。我们将进行一项多地点研究，其中疑似癫痫患者将通过正常护理途径，但在一半的病例中，我们将额外提供BioEP风险评分。我们将比较每一组的有效诊断时间、诊断类型和临床医生对所作决定的信心。这将使我们能够了解BioEP对诊断时间、决策的准确性和客观性以及对决策的信心等关键临床结果的影响。

Our proposal is focussed on Epilepsy: a serious neurological condition. Epilepsy is characterised by the tendency to have repeated, spontaneous, seizures. Since seizures occur seemingly at random, epilepsy diagnosis is very challenging. The electroencephalogram (EEG) is an important clinical tool for revealing abnormalities that indicate epilepsy. However, around 70% of recorded EEGs provide no clinical indication of epilepsy. Consequently, diagnostic delays regularly exceed a year and the rate of misdiagnosis is estimated at 30%. The cost of this misdiagnosis has been estimated at over £300M annually to the NHS.To address this challenge, Neuronostics have developed a computer Biomarker ofEpilepsy: BioEP. BioEP works by integrating routinely acquired EEG into a computer model of the brain. Simulations of this model, interrogated using advanced machine learning algorithms, provide a risk score showing how easy it is for seizures to occur. For EEGs currently considered clinically non-contributory, the BioEP risk score provides valuable additional diagnostic information. Research underpinning BioEP has been recognised by the Lancet Neurology (as a breakthrough study in epilepsy in 2015), the Epilepsy Foundation (through a commercialisation prize in 2017) and Nature (Neuronostics was one of the final four in the SpinOff Prize 2021).Through Innovate UK and NIHR funding, Neuronostics have developed a prototype of a diagnostic decision support tool powered by BioEP, that enables semi-automated feature extraction and report writing. Neuronostics have recently completed a retrospective study (using data from people with a known and accurate diagnosis) to validate the performance of the BioEP. The largest study of its kind, we demonstrated that BioEP can reveal susceptibility to epilepsy in apparently normal EEGs at a level that out performs current clinical practice.In the current project, we will undertake a prospective study to demonstrate the clinical utility of the BioEP risk score. We will conduct a multi-site study where people with suspected epilepsy will go through the normal care pathway, but in half the cases we will additionally provide the BioEP risk score. We will compare the time to working diagnosis, the types of diagnosis made, and the confidence of the clinician in the decision made in each arm. This will enable us to understand the impact BioEP has on the key clinical outcomes of time to diagnosis, accuracy and objectivity of decision making, as well as confidence in the decision reached.