PFI:BIC Human-Centered Smart-Integration of Mobile Imaging and Sensing Tools with Machine Learning for Ubiquitous Quantification of Waterborne and Airborne Nanoparticles

这种创新伙伴关系：建设创新能力（PFI：BIC）项目的重点是创建一个以人为本的智能工具集和服务系统，用于现场和无处不在的量化和纳米尺寸物体的自动表征/分类。纳米颗粒被用于越来越多的商业和工业产品，但其对健康和环境的影响仍在争论之中。纳米材料的毒性不仅因材料不同而异，而且高度依赖于暴露剂量。在通过长期毒理学研究充分了解纳米颗粒的毒性效应之前，开发一种灵敏的方法来检测纳米颗粒在环境和日常生活中的释放和时空分布是一个高度优先事项。尽管迫切需要对纳米颗粒分布进行广泛的检测和量化，但目前的技术缺乏用于纳米颗粒污染的普遍且具有成本效益的映射和量化的适当特征。该项目旨在创建一个变革性的以人为本的工具集，用于对房屋，工作场所和环境中发现的纳米材料进行现场和无处不在的量化和自动表征，其基础是计算成像和移动的传感技术与基于大数据的动态机器学习算法的成本效益整合。该项目的核心挑战是将目前用于纳米颗粒定量和表征的庞大而昂贵的实验室设备转化为现场便携，易于使用，具有成本效益和快速的分析设备和智能服务系统，旨在被消费者在日常生活中大量使用。为了解决这一挑战，将基于嵌入移动的手机中的大规模生产的互补金属氧化物半导体（CMOS）传感器芯片开发高灵敏度的光学成像系统，该传感器芯片具有非凡的信噪比（SNR）和大视场，用于基于高通量机器学习的自动化纳米颗粒分析和分类。其中一种方法是将联合收割机与纳米颗粒周围的自组装纳米透镜相结合，从而显着增强成像SNR和对比度。这种方法的目的是实现单个纳米颗粒、单分散样品和复杂多分散混合物的自动检测和尺寸测量，其中样品浓度可以跨越约5个数量级，颗粒尺寸可以从40 nm到毫米级，与现有的纳米颗粒尺寸测量方法相比，这提供了无与伦比的性能指标。将实施的另一种方法是开发用于分布式纳米颗粒成像和感测的基于高灵敏度多模（例如荧光加暗场）移动的手机的显微镜平台。此外，在大数据分析和机器学习工具方面，该项目中的技术可以自适应地学习“语义”相似性，可用于更准确的数据分类。这些技术与文献中迄今为止开发的现有技术不同。 现有的技术仅基于信号相似性，这在多模态数据上不能很好地工作。该项目的智能和自适应方法是文献中第一个带有置信界限的方法，即它们不仅具有准确分类信息的能力，而且还提供了关于这种分类的准确性的保证，这对于自学习智能服务系统非常重要。通过这些与基于自适应大数据的决策分析和量化算法集成的现场便携式设备，将创建各种消费者样品中的纳米颗粒浓度和尺寸分布的时空图，用于公共或个人监测（例如，在家中、工作场所的水载/气载颗粒或沿着高速公路的气载颗粒等的测量）。这项变革性研究的更广泛影响包括：（1）这些纳米粒子传感和量化平台以及智能服务系统的发展将扩展当前光学计量科学的边界，从而在纳米光子学和光学显微镜领域取得新进展;（2）这些设备也将易于转化为各种生物医学，化学和材料科学应用，这将对消费市场和相关产品中纳米技术的使用和监管产生重大影响。(3)该项目将通过对纳米颗粒污染和暴露的无处不在的量化和时空映射/监测，甚至在非实验室环境中，提供范式转变，帮助揭示和更好地理解消费者层面的各种因果关系，由于现有的纳米成像，检测和量化技术的限制，迄今为止尚未确定，还提供潜在健康风险的地图。(4)该项目还将在加州建立一个补充性的教育推广计划。该跨组织跨学科项目的牵头机构和主要合作伙伴包括：牵头学术机构：加州大学，洛杉矶，加州，工程学院，电气和生物工程系;主要工业合作伙伴：Holomic LLC（位于加利福尼亚州洛杉矶的小型企业）;其他行业合作伙伴：Google Inc.（位于加利福尼亚州山景的大型企业）。

This Partnerships for Innovation: Building Innovation Capacity (PFI:BIC) project focuses on the creation of a human-centered smart toolset and service system for on-site and ubiquitous quantification and automated charaterization/classification of nanosize objects. Nanoparticles are being used in more and more commercial and industrial products while their health and environmental implications are still under debate. The toxicity of nanomaterials not only varies among different materials, but is also highly dependent on the dose of exposure. Developing a sensitive method to detect the release and spatio-temporal distribution of nanoparticles in the environment as well as in daily lives is a high priority before their toxicity effects are fully understood via long-term toxicological studies. Despite this urgent need for widespread detection and quantification of nanoparticle distributions, current technologies are lacking appropriate features for ubiquitous and cost-effective mapping and quantification of nanoparticle contamination. This project aims to create a transformative and human-centered toolset for on-site and ubiquitous quantification and automated characterization of nanomaterials found in houses, workplaces and the environment based on the cost-effective integration of computational imaging and mobile sensing techniques with big data based dynamic machine learning algorithms. The central challenge in this project is to translate the bulky and expensive laboratory equipment currently used for nanoparticle quantification and characterization to field-portable, easy-to-use, cost-effective, and rapid analysis devices and smart service systems aiming to be massively used by consumers in their daily routines. To solve this challenge, highly sensitive optical imaging systems will be developed based on mass-produced Complementary Metal-Oxide Semiconductor (CMOS) sensor chips embedded in mobile phones with extraordinary signal to noise ratios (SNR) and large fields-of-view for high-throughput machine learning based automated nanoparticle analysis and classification. One approach this will take is to combine computational microscopy with self-assembled nanolenses around nanoparticles that significantly enhance imaging SNR and contrast. The aim of this approach is to enable automated detection and sizing of individual nanoparticles, mono-dispersed samples, and complex poly-dispersed mixtures, where the sample concentrations can span ~5 orders-of-magnitude and particle sizes can range from 40 nm to millimeter-scale, which provide unmatched performance metrics compared to existing nanoparticle sizing approaches. Another approach that will be implemented is the development of highly sensitive multi-modal (e.g. fluorescence plus dark-field) mobile phone based microscopy platforms for distributed nanoparticle imaging and sensing. Furthermore, in terms of big data analysis and machine learning tools, the techniques in this project can adaptively learn "semantic" similarities that can be used for more accurate data classification. These techniques are unlike existing techniques developed so far in the literature. The extant technologies are based only on signal similarities, which do not work well on multi-modality data. The smart and adaptive methods of this project are the first in the literature that come with confidence bounds, that is, they not only have the capability to accurately classify the information, but they also provide guarantees about the accuracy of this classification, which is quite important for self-learning smart service systems. Through these field-portable devices that are integrated with adaptive big data based decision analytics and quantification algorithms, spatio-temporal maps of nanoparticle concentrations and size distributions in various consumer samples will be created for public or personal monitoring (e.g., measurements of waterborne/airborne particles at home, workplace, or airborne particles along a freeway, etc.).The broader impacts of this transformative research include (1) The development of these nanoparticle sensing and quantification platforms and smart service systems will extend the boundaries of current optical metrology science, resulting in new advances in the fields of nanophotonics and optical microscopy (2) These devices will also be easy to translate into various biomedical, chemical and material science applications, significantly impacting the use and regulations of nanotechnologies in consumer market and related products. (3) This project would deliver a paradigm-shift by ubiquitous quantification and spatiotemporal mapping/monitoring of nanoparticle contamination and exposure even in non-laboratory settings, assisting in the revelation and better understanding of various cause-effect relationships at the consumer level that have remained unidentified so far due to the limitations of existing nano-imaging, detection and quantification technologies, also providing maps of potential health risks. (4) This project will also establish a complementary educational outreach program based in California.The lead institution and primary partners included in this cross-organizational interdisciplinary project are: Lead Academic Institution: University of California, Los Angeles, CA, School of Engineering, Electrical and Bioengineering Departments; Primary Industrial Partner: Holomic LLC (Small Business located in Los Angeles, CA); Other Industrial Partner: Google Inc. (Large Business located in Mountain View, CA).