Noncontact Remote Monitoring for the Detection of Opioid-Induced Respiratory Depression

美国阿片类药物危机继续对人类生活和正在进行的新冠肺炎造成灾难性影响 大流行正在加剧其影响。根据美国疾病控制与预防中心公布的统计数据，91,799人吸毒过量 死亡发生在2020年的美国，经年龄调整的过量死亡人数比2019年增加了31%， 2020年。此外，导致呼吸抑制的阿片类药物占所有药物过量的75%。 美国的死亡人数。我们建议在非侵入性生命体征监测工作的基础上开发FDA- 监管医疗设备在阿片类药物诱导呼吸监测中的初步应用 抑郁症。这包括对正在接受大剂量阿片类药物治疗的慢性疼痛患者进行家庭监测。 处方药或阿片类药物使用障碍(OUD)患者以及监测对象 在监督注射地点(也称为监督消费空间)使用OUD。我们的总体目标是 非接触式多模式阿片类药物呼吸监测系统的研制 在家里和有监督的注射地点出现抑郁。虽然雷达能够穿透衣服 和毯子测量呼吸引起的胸壁运动，它需要深度的指导 成像以瞄准一个人和胸部区域。我们的具体目标是：1.使用 非接触式监测系统。我们目前的技术能够以较高的速度检测呼吸频率 固定对象的准确度。然而，呼吸抑制的可靠检测涉及到 监测呼吸频率、模式和深度(即潮气量)。作为这一具体目标的一部分，我们将制定 一种使用雷达和深度信息估计静止物体潮气量的框架，其中我们 根据胸壁位移估计潮气量。此外，我们将提取特征来表征 从获取的雷达信号中提取呼吸模式。作为对此评估框架的初步验证，我们的 该系统将在20名健康志愿者身上进行测试。测试结果将为我们提供初步数据 关于雷达和基于深度的潮气量估算与黄金相比的准确性 标准的。2.开发和验证集成传感器数据以检测呼吸的框架 抑郁症。在这个特定的目标中，我们将开发一个框架来使用呼吸频率、呼吸模式、 以及来自雷达和深度相机的潮气量信息来确定呼吸抑制是否有 发生了。这涉及两个步骤的方法，其中我们提取呼吸特征来表征呼吸 模式来补充呼吸频率和潮气量，然后使用机器学习模型来检测 发生呼吸抑制。为了帮助设计正确的模型，我们将使用我们的雷达收集数据 以及麻醉猪经历阿片类药物呼吸抑制的深度成像系统。

The US opioid crisis continues to have a catastrophic impact on human lives and the ongoing COVID-19 pandemic is compounding its effects. Based on the statistics published by the CDC, 91,799 drug overdose deaths occurred in the US in 2020, where the age-adjusted overdose deaths increased by 31% from 2019 to 2020. In addition, opioids, which cause respiratory depression, were involved in 75% of all drug overdose deaths in the US. We propose to build on our work in non-invasive monitoring of vital signs to develop an FDA- regulated medical device with a primary application in monitoring patients for opioid-induced respiratory depression. This includes at-home monitoring of patients with chronic pain being treated with high-dose opioid prescription medications or patients suffering from opioid use disorder (OUD) as well as monitoring subjects with OUD at supervised injection sites (also known as supervised consumption spaces). Our overall goal is to develop a non-contact multi-modal monitoring system for the detection of opioid-induced respiratory depression at home and in supervised injection sites. While radar is capable of penetrating through clothing and blankets to measure chest wall movements resulting from respiration, it requires the guidance of depth imaging to target a person and the chest area. Our specific aims are: 1. Estimate tidal volume using a noncontact monitoring system. Our current technology is capable of detecting respiratory rate with a high degree of accuracy for stationary subjects. However, robust detection of respiratory depression involves monitoring of respiratory rate, pattern, and depth (i.e., tidal volume). As part of this specific aim, we will develop a framework to estimate tidal volume of a stationary subject using radar and depth information, where we estimate tidal volume from chest wall displacements. Furthermore, we will extract features to characterize respiratory pattern from the acquired radar signal. As a primary validation of this estimation framework, our system will be tested on 20 healthy volunteers. The outcome of the test will provide us with preliminary data regarding the accuracy of the radar and the depth-based tidal volume estimation as compared with the gold standard. 2. Develop and validate a framework for integrating data from sensors to detect respiratory depression. In this specific aim, we will develop a framework to use the respiratory rate, respiratory pattern, and tidal volume information from the radar and depth camera to determine if respiratory depression has occurred. This involves a two-step approach, where we extract respiratory features to characterize respiratory patterns to complement respiratory rate and tidal volume, and then use a machine learning model to detect the occurrence of respiratory depression. To help with design the right model, we will collect data using our radar and depth imaging system from anesthetized pigs going through opioid-induced respiratory depression.