Development of a Novel Diagnostic Test for Pulmonary Embolism Based on Artificial Intelligence and Spectral Analysis of Blood

基于人工智能和血液光谱分析的新型肺栓塞诊断测试的开发

• 批准号: 10081921
• 负责人: Jeremy Ford Huckins
• 金额: $ 29.97万
• 依托单位: BIOCOGNIV INC.
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-09-01 至 2022-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10081921
• 关键词: Accident and Emergency department; Acute; Address; Adult; Affect; Age; Angiography; Area; Artificial Intelligence; Autopsy; Biological Assay; Biological Markers; Blood; Blood Proteins; Blood Tests; Blood coagulation; Cardiovascular Diseases; Cardiovascular system; Cessation of life; Chest Pain; Creatine Kinase MB Isoenzyme; Data; Data Set; Decision Making; Deep Vein Thrombosis; Development; Diagnosis; Diagnostic; Diagnostic tests; Dyspnea; Emergency Department patient; Evaluation; Fibrin fragment D; Goals; Gold; Hospitals; Image; Immunoassay; Intravenous; Ionizing radiation; Life; Lower Extremity; Lung; Machine Learning; Measures; Medical center; Methodology; Methods; Modeling; Multicenter Studies; Myocardial Infarction; Myoglobin; Patient Care; Patient Triage; Patients; Performance; Perfusion; Phase; Physicians; Plasma; Population; Predictive Value; Process; Pulmonary Embolism; ROC Curve; Radiation; Risk; Sampling; Scanning; Seizures; Sensitivity and Specificity; Shortness of Breath; Specificity; Standardization; Symptoms; Syncope; Test Result; Testing; Training; Travel; Troponin I; Universities; Validation; Venous system; Vermont; X-Ray Computed Tomography; base; clinical care; cost; cost effective; data modeling; improved; inflammatory marker; instrumentation; molecular marker; novel; novel diagnostics; patient safety; performance tests; point of care; standard of care; targeted biomarker; ventilation

Pulmonary Embolism (PE) is a potentially life-threatening condition that affects adults of all ages yet can present with a myriad of symptoms, ranging from chest pain to shortness of breath, syncope, or seizure. Currently, physicians can assess for this condition with a blood test (D-dimer) which has high sensitivity but very poor specificity, thus resulting in a larger number of false positives. Alternatively, or in the case of a positive D-dimer, computed tomography pulmonary angiography (CTPA) or ventilation-perfusion (VQ) scan can be used, both of which are expensive and expose the patient to a significant amount of ionizing radiation. To develop a more specific blood test for PE, Biocogniv will apply state-of-the-art artificial intelligence (AI) to aggregate analysis of existing blood biomarkers measured on two commercially available multiplex immunoassay platforms, one of which is currently used in hospital labs. A sufficiently accurate, rapid and cost-effective test could be broadly applied (i.e., like troponin-I for myocardial infarction) to reduce overuse of CT, simplify ED decision making, and reduce the number of deaths from unrecognized PE. In this proposed Phase I single center study, Biocogniv will collaborate with the University of Vermont Medical Center (UVMMC) to demonstrate proof-of-concept for diagnosing PE in emergency department (ED) patients for whom there is sufficient concern for PE to warrant a D-dimer test as part of routine clinical care. Specific Aim I is to collect blood from 225 emergency department (ED) patients at UVMMC who were suspected of having PE (including 75 patients that are confirmed to have PE by CTPA), and analyze the blood plasma with quantitative immunoassays to create training and validation datasets for AI analysis. Immunoassays will be comprised of a set of 6 rapid FDA-cleared chemiluminescent immunoassays performed in parallel, and a 20-plex bead-based immunoassay panel targeting known cardiovascular and inflammatory markers associated with acute PE. Specific Aim II is to develop AI data models for two pretest populations—all D-dimer tested patients and just D-dimer positive patients—for bead-based and point-of-care immunoassay datasets (analyzed separately), then evaluate the performance of each data model on a subset of blood plasma data withheld for validation. As part of the evaluation of AI data model performance, the potential impact of study size on data model accuracy will be simulated by plotting specificity as a function of training sample number to show that adding more samples can improve test results. The performance of each approach (i.e., AI methodology applied to a given immunoassay panel for a given pretest population) will be compared and used to plan a Phase II multi-center study and to identify and attract a suitable instrumentation partner. Biocogniv’s end goal is to develop a rapid, highly specific and sensitive FDA cleared test for PE that will become the standard of care for initial diagnosis and triage of patients with potential pulmonary embolism, disrupting the $2B/year D-Dimer market and improving ED patient care.

肺栓塞（PE）是一种可能危及生命的疾病，影响所有年龄段的成年人， 表现出从胸痛到呼吸急促、晕厥或癫痫发作的各种症状。 目前，医生可以通过血液检测（D-二聚体）来评估这种情况，该检测具有高灵敏度， 非常差的特异性，从而导致大量的假阳性。或者，或者在 D-二聚体阳性、计算机断层扫描肺血管造影（CTPA）或通气-灌注（VQ）扫描可 这两种方法都是昂贵的，并且使患者暴露于大量的电离辐射。到 为PE开发更具体的血液测试，Biocogniv将应用最先进的人工智能（AI）， 对两种市售多重检测仪测量的现有血液生物标志物进行汇总分析 免疫分析平台，其中一个目前在医院实验室使用。一个足够准确、快速和 成本有效的测试可以被广泛应用（即，如肌钙蛋白-I用于心肌梗死），以减少过度使用 CT，简化艾德决策，减少未被识别的PE死亡人数。 在这项拟议的I期单中心研究中，Biocogniv将与佛蒙特大学医学院合作， 中心（UVMMC）证明诊断急诊科（艾德）患者PE的概念验证 对于那些有足够的PE关注，以保证D-二聚体测试作为常规临床护理的一部分。具体 目的I是收集225名UVMMC急诊科（艾德）疑似 PE患者（包括经CTPA证实为PE的75例患者），并使用 定量免疫测定，为AI分析创建训练和验证数据集。免疫测定将 包括一组平行进行的6个FDA批准的快速荧光免疫测定， 20-靶向已知心血管和炎症标志物的基于复合珠的免疫测定组 急性PE具体目标II是为两个预测试人群开发AI数据模型-所有D-二聚体测试 患者和仅D-二聚体阳性患者-用于基于微珠和床旁免疫测定数据集 （单独分析），然后评估每个数据模型在血浆数据子集上的性能 保留以供验证。作为AI数据模型性能评估的一部分，研究的潜在影响 将通过绘制特异性与训练样本数量的函数关系图来模拟数据模型准确度的大小， 表明增加更多的样品可以改善测试结果。每种方法的性能（即，AI 用于给定预试验人群的给定免疫测定组的方法学）进行比较和使用 计划II期多中心研究，并确定和吸引合适的仪器合作伙伴。 Biocogniv的最终目标是开发一种快速，高度特异性和灵敏度的FDA批准的PE测试， 成为潜在肺栓塞患者初步诊断和分诊的标准护理， 扰乱每年20亿美元的D-二聚体市场，改善艾德患者护理。