High performance wearable body odor sensor arrays for disease detection and monitoring

项目摘要 许多疾病，无论是内部和皮肤，有不同的气味与他们有关，他们的识别 可以提供独特的诊断线索，指导实验室评估，并促进和加快治疗。电流 体臭分析依赖于台式仪器，但它们太笨重，无法在护理点、家庭或 职场电子鼻技术为体臭分析提供了一种简单、轻便和低成本的替代方案，但它们 对环境变化高度敏感（例如，温度和湿度）。此外，电子鼻还受到 当暴露于约100种皮肤散发的蒸汽分析物时， 同步这些缺点使得电子鼻模式识别困难且不准确。为了克服这一点， 我们建议开发一种可穿戴的微型气相色谱（GC）设备， 纳米电子鼻和生命体征传感器，并使用它来分析身体气味相关的>20 疾病/状况。在这种可穿戴设备中，皮肤散发的蒸汽将通过微型GC进行预分离， 通过嵌入在GC柱末端的石墨烯电子鼻检测以生成时间序列图案。 由于蒸汽分析物将一次洗脱出一个或几个，因此电子鼻的模式识别将非常重要。 更简单更准确。温度/湿度问题也将大大减少，因为蒸汽 传感器对温度变化不敏感。此外，GC中的预浓缩器是疏水的， 不会截留水，剩余的水将通过GC从其他蒸汽中分离出来。最后 由于预浓缩，GC柱内的蒸气浓度比皮肤表面附近高> 50倍 效果由于这些优点，模式识别和疾病检测能力将显著提高。 增强我们的多学科团队在生物医学/电气工程、数据和通信技术方面拥有所需的专业知识。 科学，以及各种临床领域，包括皮肤科，急诊医学，肺病学， 儿科.我们将实现以下具体目标。目标1.开发和制造可穿戴设备， 一次性用品。我们将建造20个集成石墨烯电子鼻的自主可穿戴GC设备。的 可穿戴设备将是小的、轻的（~200 g）、电池供电的。我们还将制作2,000个定制的 一次性塑料蒸汽采样室，采用注塑成型，并配有生命体征传感器。目标二。 开发和实施算法来分析时间序列模式。我们将根据以下内容开发算法： 深度学习来分析时间序列模式和生命体征数据。我们将训练自动编码器神经网络 模型并将其应用于参与者的功能。将训练正则化分类模型以识别 积极的患者。Shapely值将用于解释模型做出的预测。 目标3：分析>20种疾病/状况。我们将从密歇根大学健康中心招募患者 系统，然后使用目标1和2中开发的可穿戴设备和算法来分析>20 在四个不同的专业：皮肤科，急性护理，肺内科和儿科疾病/条件。

Project Summary Many diseases, both internal and cutaneous, have distinct odors associated with them, and their identification can provide unique diagnostic clues, guide laboratory evaluation, and facilitate and expedite treatment. Current body odor analysis relies on benchtop instruments, but they are too bulky for use at point-of-care, home or workplace. E-nose technologies provide a simple, light, and low cost alternative for body odor analysis, but they are highly susceptible to environmental changes (e.g., temperature and humidity). Additionally, e-nose suffers from strong cross-talk among the sensing elements when it is exposed to ~100 skin-emitted vapor analytes simultaneously. These drawbacks make e-nose pattern recognition difficult and inaccurate. To overcome this, we propose to develop a wearable micro-gas chromatography (GC) device integrated with graphene based nano-electronic e-nose and vital sign sensors, and use it to analyze body odors related to >20 diseases/conditions. In this wearable device, skin-emitted vapors will be pre-separated by micro-GC and then detected by the graphene e-nose embedded at the end of the GC column to generate time-series patterns. Because vapor analytes will be eluted out one or a few at a time, pattern recognition by e-nose will be much simpler and more accurate. The temperature/moisture issues will also be greatly reduced since the vapor sensors are insensitive to temperature changes. Additionally, the pre-concentrator in the GC is hydrophobic and does not trap water, and the remaining water will be separated out from other vapors through GC. Finally, the vapor concentration inside the GC column is >50X higher than near the skin surface due to the pre-concentration effect. Because of these advantages, the pattern recognition and disease detection capability will be significantly enhanced. Our multidisciplinary team has the needed expertise in biomedical/electrical engineering, data science, and a variety of clinical realms including dermatology, emergency medicine, pulmonology, and pediatrics. We will achieve the following specific aims. Aim 1. Develop and fabricate wearable devices and disposables. We will build 20 autonomous wearable GC devices integrated with graphene e-nose. The wearable device will be small, lightweight (~200 g), battery-powered. We will also fabricate 2,000 customized disposable plastic vapor sampling chambers using injection molding with vital sign sensors incorporated. Aim 2. Develop and implement algorithms to analyze time-series patterns. We will develop the algorithm based on deep learning to analyze time-series patterns and the vital sign data. We will train an autoencoder neural network model and apply it to the features from participants. A regularized classification model will be trained to identify the positive patients. Shapely values will be used to provide explanations for the prediction that the model makes. Aim 3. Analyze >20 diseases/conditions. We will recruit patients from the University of Michigan Health System and then use the wearable devices and algorithms developed in Aims 1 and 2 to analyze >20 diseases/conditions in four different specialties: dermatology, acute care, pulmonary medicine, and pediatrics.