自动驾驶车载成像仪
自动驾驶车载成像仪
(“智能析体（CT）机器人”)
王革
二零一八年六月二十四日于美国纽约阿尔伯尼
小编的话：本期迎来了范的导师王革教授的投稿，这是基于范同学和导师的君子协定（范写两篇杂志文章，王老师就支持一篇公众号文章）。愿大家常在“纯真学者出神入化”公众号上看到王老师的投稿，哈哈。今天王老师将与我们漫谈未来医疗成像技术。中文版本在先，英文原稿随后。 两相对照，相得益彰。
要点一：文明源于联系和移动。铁路和公交车站虽在旅行上发挥重要作用，但如今私家汽车更为流行。近来，优步出行风头强健，自动驾驶汽车也显露端倪。虽然我们现在主要在超市和商场购物，但网络购物和快递的趋势已经明朗。不久之前，我们还习惯于去电影院和剧院，后来我们享用家庭影院。而如今我们大部分时间用手机看电影电视。很多人相信区块链技术将取代银行，促进社会革命。总结来说，在信息互联（互联网）和万物互联（物联网）之后，服务互联（“服联网”）会成为下一波浪潮。信息，产品和服务的个性化和最优化利用需要以泛网络化， 去中心化，机器学习和人工智能为前提， 从而改善民生，促进民主和保障民权。
要点二：影像信息独特且重要。视觉是人类获取信息的主要方式之一。在美国国立卫生院的所有研究所中，唯一不和人类器官和疾病直接联系的就是它的生物影像和生物工程研究所。医学影像是数亿美元的产业，已经带来巨大的福利，正在经历深刻的变革，未来必将产生更大的社会影响。
要点三：医疗影像的未来。我们预见到医疗影像的范式转变：从以医院为中心到智能化地将医疗成像服务随时投递到需要的地方。智能化的自动驾驶车载成像仪是大家需要的，尤其是自然灾害地区，恐怖袭击和战争地带。由于其低成本，自动化及便捷性，这种设备也可用在欠发达地区日常的癌症普查及诊断。
要点四：高新科技的汇聚。由于过去十余年在工程领域前所未有的进展，自动驾驶车载成像仪可以集结尖端的影像技术，机器学习，机器人技术，高性能计算机，自动驾驶技术，进而改变影像世界的面貌。去年我们提出机器学习会在五年内改革放射学科：https://aapm.onlinelibrary.wiley.com/doi/pdf/10.1002/mp.12204。许多人也相信人类将在五年之内拥有可以上路的自动驾驶汽车： https://www.quora.com/When-will-self-driving-cars-be-available-to-consumers。同时，以波士顿动力为代表的机器人技术也有很大发展： https://www.bostondynamics.com。这些快速进步的技术可以被结合起来以成就我们设想的智能成像原型机。
要点五：工程研发中心。这种原型机可以通过一个工程研发中心打造以用于疾病筛查，灾害救援和战场服务。对于原型机来说，最开始的成像方式可以是CT,超声或者光学成像，核磁共振等其它成像方法随着项目进展可以再加入进来。这种机器人将具备重要的自主成像能力同时预期成本不高。自动驾驶汽车，无人机，行走的机器人都可以是潜在的成像设备的载体。
要点六：研发可行性。基于CT的原型机将会依照我们几年前设计的方案研发：http://live.iop-pp01.agh.sleek.net/2015/01/28/how-to-create-a-low-cost-ct-scanner。 这个方案的成本估算就已经比现在的医用CT扫描机便宜很多。两年多以前，我们组开始研究机器学习在医疗成像和“端到端”流程中的应用：https://ieeexplore.ieee.org/document/7733110/?reload=true，这项研究获得了通用电气公司全球研发中心的资助和协作。在和我们伙伴的合作过程中，我们的设计和方法将会不断得到改善。现在科技进步神速，包括X光硬件(低能耗轻量X光管和集成高压发生器，高分辨率X光计数检测器)，先进算法(低维流形，机器学习)，智慧材料，计算机视觉，自动语言翻译，互联网和高性能计算。在此基础上，这种原型机（即自动驾驶车载成像仪 或“智能析体（CT）机器人”)，的研发必将有广阔的应用前景。
英文原稿写于本年四月十八日，欢迎切磋，共同进步。
Auto-driving Vehicle-basedAffordable Tomography-Analytics Robots (AVATAR)
Civilization Spanned by Mobility & Connectivity– Rail/bus stations made milestones in the transportation history but quickly overwhelmed by privately-ownedcars. Now, Uber taxi become popular, and auto-driving cars are on the horizon. Similarly, supermarkets/malls are where we shop but the trend is moving towardsinternet-shopping and door-to-door delivery. Not long ago, we used to go to cinemas and theaters for entertainment. Then, we built home theaters. Today, we use smart phones to watch TVs most of time. Yet, many believe that theblockchain technology will revolutionize the society; say, eliminating banks. In summary, after “Internet of information”,“Internet of things” are under activedevelopment, and “Internet of service”seems the next wave. The individualized and optimized use of information,products, services demands decentralization, interconnection, and machineintelligence, promoting democracy and improving quality of life.
Importance of Medical Imaging – Human vision is a primary source of information. Among all NIH institutes, the only institute not directly related to human organs and diseases is NIBIB, an imaging/BMEinstitute. Medical imaging is a multi-billion dollar business, brings tremendous healthcare benefits, undergoes dramatic transformation, and promisesa major societal impact.
Future of Medical Imaging – We envision a paradigm shift in medical imaging, from hospital/clinic/center-oriented to mobile, intelligent, and integrated services promptly delivered wherever and whenever needed. The Auto-driving Vehicle-based Affordable Tomography-Analytics Robots(AVATAR) are most desirable on natural disaster spots, after terrorists’ attacks, and near battle fields. Also, AVATAR are advantageous in routine healthcare imaging such as cancer screening because of potentially much-reduced cost, full automation, and greatly-improved convenience, especially for rural areas and under-developed countries.
Convergence of Maturing Engineering High-techs–Given unprecedented progresses in the engineering field over the past decade or so, AVATAR is timely to integrate cutting-edge medical imaging, machine learning, robot, high-performance computing, internet, and auto-driving technologies, and change the landscape of the imaging world. In the last year, we argued that “Machine learning willtransform radiology significantly within the next 5 years” (https://aapm.onlinelibrary.wiley.com/doi/pdf/10.1002/mp.12204). Coincidentally, with several good reasons many people believe that you will have autonomous cars within 5 years (https://www.quora.com/When-will-self-driving-cars-be-available-to-consumers). Also, Boston Dynamics is famous for the development of quadruped robots withamazing performance in battle fields (https://www.bostondynamics.com), supported by DARPA and DI-Guy. These and other techniques are rapidly evolving,and can be now combined to make AVATAR prototypes.
ERC Center for AVATAR –It is possible to build an ERC center is to developAVATAR prototypes as testbeds for cancer screening and global disaster relief (incollaboration with Massachusetts General Hospital), as well as military traumacare near battle fields. The initial imaging modality for AVATAR will be x-ray computed tomography (CT), and other imaging modalities such as ultrasound andoptical imaging, MRI and nuclear imaging will be added as the project goes on. AVATAR will implement critical imaging capabilities previously unavailableand/or achieve a fraction of the total cost of today’s corresponding imagingprocedures. Auto-driving cars, autonomous drones and walking robots can be thevehicles of interest.
Preliminary Studies Related to AVATAR – An AVATARCT scanner will be developed in reference to the prototype we designed severalyears ago, http://live.iop-pp01.agh.sleek.net/2015/01/28/how-to-create-a-low-cost-ct-scanner, which is already an order ofmagnitude cheaper than the current commercial products. More than two yearsago, our group started working on machine learning for medical imagingespecially tomographic reconstruction and end-to-end analysis (https://ieeexplore.ieee.org/document/7733110/?reload=true), recently funded by GE GlobalResearch Center (GRC). Our existing design and methods will be improved incollaboration with GRC and other partners, based on rapidly-evolving low-power light-weightx-ray tubes and integrated generators, high-resolution and photon-countingdetectors, low-dimensional manifold and data-driven machine learning, smartmaterials, computer vision, language translation, high-performance computing,and robotic technologies, for a good portion of which RPI has world-classexpertise on campus.
作者简介：
GE WANG is a Clark and Crossan Chair Professor and the Director of the Biomedical Imaging Center, Rensselaer Polytechnic Institute, USA. He authored the papers on the first spiral/helical cone-beam/multi-slice CT algorithm (http://iopscience.iop.org/article/10.1088/0031-9155/52/6/R01/meta). Currently, there are over 100 million medical CT scans yearly with a majority in the spiral cone-beam mode. He and his collaborators pioneered bioluminescence tomography. His group published the first papers on interior tomography (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3775479) and omnitomography (all-in-one) to acquire diverse data sets simultaneously (all-at-once) (https://aapm.onlinelibrary.wiley.com/doi/full/10.1118/1.4929559). His results were featured in Nature, Science, and PNAS (https://www.nature.com/polopoly_fs/1.9645!/import/pdf/480303a.pdf), and recognized with awards. He has written more than 430 peer-reviewed journal publications, including the first perspective on deep tomographic reconstruction/imaging (https://ieeexplore.ieee.org/document/7733110). As the lead guest editor, he co-edited with his peers five special issues for IEEE Trans. Medical Imaging, which is the flagship journal in the medicla imaging field. He is a Fellow of SPIE, OSA, AIMBE, AAPM, and AAAS.
想关注更多有趣的机器学习/计算物理/医疗影像的科普/前沿研究，请关注我们的公众号：
这是一群致力于科研传播的faculty & PhD记录分享点滴思考的平台，这里有我们在各自领域（机器学习，医疗影像，材料科学，凝聚态物理）涉猎研究的点滴感悟，有我们在国内，海外求学工作奋斗的酸甜苦辣，亦有偶尔的风月和我们的诗与远方。