中国开源软件网

当前位置: 首页 > 科学 >

李开复连线杂志专栏,将加速AI赋能医疗的场景落地

时间:2020-05-22 22:34来源:互联网 作者:小狐

李开复连线杂志专栏,将加速AI赋能医疗的场景落地(图1)

2020年元旦前夜,一家位于加拿大多伦多市的人工智能(AI)企业BlueDot捕捉到一些异常:中国武汉市海鲜市场周边出现多起罕见肺炎病例。BlueDot迅即反应,运用自然语言处理、机器学习等技术,结合大数据和定位追踪,迅速向合作的政府部门和公共卫生机构客户传送警报并报告扩散状况。

BlueDot所监测到的异状,正是数月后撼动全球的新型冠状病毒肺炎(Covid-19)这比世界卫生组织首度公开警示新冠病毒的时间还要早上9天。

BlueDot的AI平台示范了人工智能技术对重大疫情能起到早期预警的功用,过去几个月里,AI在这场全球抗疫战的许多方面发挥了独特作用:从疫情预测到筛检,从接触警示到快速诊断,从前线无人配送到实验室药物研发,人工智能助力防疫派上了不少用场,为特定场景应用赋能。

随着疫情在全球蔓延,AI技术的创新应用也在各地相继落地。

在韩国,基于地理位置的信息传递已经成为控制病毒传播的重要工具,当人们靠近确诊病例时,就会收到基于位置的紧急信息提醒。在中国,阿里巴巴推出的AI算法能够在20秒内诊断出疑似病例(比人类检测快了近60倍)准确率高达96%。无人配送车辆很快被投入到人类难以承受的场景,代替人类执行高传染风险的运输任务。湖北、广东等省份的多家医院相继使用机器人为病人或被隔离家庭运送食物、药品和物资。而在美国加州,计算机科学家正在研发能远程检测独居老人健康情况的,一旦老人出现身体异常症状,就会发出即时警报。

不过,目前 人工智能在公共健康体系的应用仍显零散也未成体系。坦率说,过去四个月内,AI在抗疫之战中的表现并不十分突出,我最多只能给它打分 “B-”新冠大流行暴露了我们的医疗的脆弱性:预警响应不充分、通报信息不精确、医疗物资分配不均、医务人员超负疲惫、医院病床紧绷、疫苗研发周期长等诸多痛点。

当然,AI的零散表现也有客观原因:医疗体系可说是现代社会各类运转体系中最为复杂、陈旧不堪且难以变通的体系;且在新冠疫情袭来之前,我们并没有真正意识到医疗体系问题的紧迫性,没有提前采取相应的技术预防措施;最为关键的是,我们缺少建构AI解决方案所需的大数据。

把目光投向未来,我看到以下两个AI赋能医疗的乐观因素。

首先,作为AI燃料的医疗大数据已被激活。举例来说,机器学习数据科学平台Kaggle组建了新冠病毒开放研究数据集,名为CORD-19。它将相关数据进行汇编,并把最新研究集中收录,汇总的格式可被机器读取和解析,以便于AI进行机器学习。至今这个数据集收录了12.8万篇包含Covid-19、冠状病毒、SARS(型肺炎)MERS(中东呼吸综合症)等关联术语的医学专业学术文章。

其次,眼下全世界的医学专家和计算机科学家都将精力集中在解决疫情问题。X大奖基金会创始人彼得·戴曼迪斯(Peter Diamandis)估计,全球现在有多达2亿名的医师、科学家、护士、技术专家和工程师投入防治冠状病毒的相关研发中,他们正在进行数以万计的实验,并以“前所未有的透明度和速度”共享信息。

在国内疫情爆发后不到一个月,阿里巴巴便推出了一种AI算法,该算法基于5000多个新冠肺炎确诊病例进行训练,并关联到治疗后续诸如肺部白色阴影缩小等的成效追踪。随后,阿里巴巴将其云端AI平台向全球医疗专业人员开源,与合作伙伴联手部署更大批量的匿名数据,推出包括疫情预测、CT影像分析、冠状病毒基因组测序等模块。

据估计,现今 全球医疗数据的规模每隔几个月就翻一倍。2019年一份覆盖19个国家AI医疗市场的研究估计,AI医疗市场的年复合增长率为41.7%,从2018年的13亿美元将增长至2025年的130亿美元,主要分布在六大领域:医院工作流程、可穿戴设备、医学影像和诊断、诊疗计划、虚拟助手、以及最重要的药物研发,新冠疫情期间浮现的种种需求,将加速AI赋能医疗的场景落地。

李开复连线杂志专栏,将加速AI赋能医疗的场景落地(图2)

在后疫情时代,我期待AI将加速融入医疗体系,赋能并推动医疗改革。其中深度学习(Deep Learning)即以一种高效方法运算海量、数据的能力,是AI结合医疗最为可期的机遇之一。深度神经网络(Deep Neural Networks)作为AI的一个子领域,已经被用于医学扫描、病理切片、眼科检查甚至结肠镜检查,以得出准确而快速的算法判读。十几年后,不少国家和地区的医疗体验在AI赋能的作用下将发生根本性改变。

AI赋能医疗,首先能简化及优化现有的医疗流程,例如医院的作业流程,保险履约的繁复流程。将AI与RPA(Robotic Process Automation 机器人流程自动化)结合,可对某项工作流程进行智能拆解及优化,进而大大提高医疗的效率,预约看诊、保险理赔及其他流程性工作都会得到效率提升。AI还能加快早期诊断信息的收录并实现自动化,AI技术所能处理的文本、语言、数字的体量,无论在数量上还是精度上都是机器级别,远非人类所及。

再上一个层次的AI赋能体现在助力新药研发、基因组测序、干细胞、CRISPR(基因)等医学突破方面, AI模型和算法应用都有其用武之地。在制药行业,研发一种新药往往需要付出高昂的投入,某次成功前必有多次付诸流水的失败试验,也连带消耗巨大的时间和金钱成本。

现在,科学家们可使用AI机器学习来模拟上千个变量,它们的复合效应会对人类细胞反应产生何种影响,这类AI新药研发的技术已被用于新冠病毒疫苗和其他疗法。创新工场所投资总部位于香港的AI药物研发公司Insilico Medicine是首批对新冠病毒快速响应的企业之一,这家公司利用生成式化学AI平台设计出新药物小分子,以复制主要病毒蛋白为靶标,早在2月5日便公布了这些小分子结构。AI为新药发明开辟了一个新时代,用人工智能技术来换取药品研发周期的时间和成本,整个制药行业势将迎来翻天覆地的变革。

不久的将来,随着医疗科学和计算机科学进一步融合,我们将进入一个全面自动化的AI时代,到时人们可以通过可穿戴设备、生物传感器、智能家居检测设备等来确保自身和家人的健康。可穿戴设备和其他物联网设备的数据质量和多样性大幅提高,将能产生一个有效的良性循环。穿越到未来,下一场疫情在大范围蔓延之前就应该能够被跟踪、追溯、拦截并消灭无踪。

或许 再过15年,许多人的家里都会有AI个人助理照料我们,帮着解决全家人的日常健康所需。机器人或者无人机会把我们的药品送上门,如果需要进行手术或者外科治疗,通常会由机器人操作,或由机器人辅助人类外科医师完成。

在我的想象里,15年后的医疗健康场景可能是这个样子的:

AI的未来

2035年一个冬季早晨,我醒来后就觉得有点儿喉咙痛。我起身去洗手间,刷牙的时候,洗手间的镜子通过红外传感器测量了我的体温。刷完牙后一分钟,我的私人AI医师助理发出了警报,显示我的唾液样本部分指数异常,并在轻微低烧。AI医师助理建议我在家进行指尖探针采血。我在泡咖啡时,医师助理返回了分析结果,判断我可能是得了这个季节正在流行的两种流感其中一种。之后,我的AI医师助理建议,如果我觉得有必要家庭医生的话,有两个时间空档可以跟她通话。通话之前,家庭医生已经收到我所有症状的详细信息。她给我开了一种减充血剂和扑热息痛,一会儿无人机会把药品送到我家门口。

当然,凡涉及到病患医疗记录,就得谈谈隐私和数据保护的关键问题。我认为,任凭有用的数据各自孤岛式的存在,不善加利用,不从中提炼有价值的信息,不用以推动社会进步,是相当不负的做法。技术产生的问题应该由技术解决。随着AI技术浪潮而出现的诸如数据保护等问题,应该由更为创新的技术方法来应对。

好是, 近年联邦学习(也被称为分布式学习)已经在数据保护上取得了显著的进展。基于联邦学习技术,患者的数据将永远不会离开所在的医疗机构、医院或个人设备等原始存储设备,机器学习模型将在独立的数据集基础上进行训练处理,再进行后续整合。联邦学习、同态加密,结合可信硬件执行环境等技术,将进一步确保数据的计算、传输、存储过程能够适配不同的隐私偏好,以因应不同国家与文化对于隐私保护的需求差异。

这次新冠肺炎疫情还验证了:整体 人类命运是共同体,人们对未来运用AI等先进技术共度难关寄予一致的期盼。过去,国际合作曾消灭了全球延烧的天花,也几乎根除了小儿症。公共卫生无国界,控制及消除流行病是个毋庸置疑的共同目标。在医学领域,每个国家都能从他国的研究基础上学习受益并携手并进,全球化的数据科学,将进一步帮助人类获取对健康和疾病最为深刻、最为全面的洞悉。

经历这次疫情,我们应清醒地意识到,要将人类医疗体系推往新的高度,着实需要倾尽全球之力。

创新工场董事长兼首席执行官,创新工场人工智能工程院院长

◀▶

Covid-19 Will Accelerate theAI Health Care Revolution

Disease diagnosis, drug discovery, robot delivery—artificial intelligence is already powering change in the pandemic’s wake. That’s only the beginning.

ON NEW YEARS Evethe artificial intelligence platform BlueDot picked up an anomaly. It registered a cluster of unusual pneumonia cases in Wuhan, China. BlueDot, based in Toronto, Canada, uses natural language processing and machine learning to track, locate, and report on infectious disease spread. It sends out its alerts to a variety of clients including health care, government, business, and public health bodies. It had spotted what would come to be known as Covid-19, nine days before the World Health Organization released its statement alerting people to the emergence of a novel coronavirus.

BlueDots role in spotting the outbreak was an early example of AI intervention. Artificial intelligence has already played a useful but fragmented role in many aspects of the global fight against the coronavirus. In the past months, AI has been used for prediction, screening, contact alerts, faster diagnosis, automated deliveries, and laboratory drug discovery.

As the pandemic has rolled around the planet, innovative applications of AI have cropped up in many different locations.

In South Korea, location-based messaging has been a crucial tool in the battle to reduce the transmission of the disease. Nine out of 10 South Koreans have been getting location-based emergency messages that alert them when they are near a confirmed case.

In China, Alibaba announced an AI algorithm that it says can diagnose suspected cases within 20 seconds (almost 60 times faster than human detection) with 96 percent accuracy. Autonomous vehicles were quickly put to use in scenarios that would have been too dangerous for humans. Robots in Chinas Hubei and Guangdong provinces delivered food, medicine, and goods to patients in hospitals or quarantined families, many of whom had lost household breadwinners to the virus. In California, computer scientists are working on systems that can remotely monitor the health of the elderly in their homes and provide alerts if they fall ill with Covid-19 or other conditions.

These snapshots of AI in action against Covid-19 provide a glimpse of what will be possible in the various aspects of health care in the future. We have a long way to go. Truth be told, AI has not had a particularly successful four months in the battle of the pandemic. I would give it a “B minus” at best. We have seen how vulnerable our health care systems are: insufficient and imprecise alert responses, inadequately distributed medical supplies, overloaded and fatigued medical staff, not enough hospital beds, and no timely treatments or cures.

Health care systems around the world—even the most advanced ones—are some of the most complicated, hierarchical, and static institutions in society. This time around, AI has been able to help in only pockets of excellence. The reasons for this are simple: Before Covid-19 struck, we did not understand the importance of these areas and act accordingly, and crucially as far as AI is concerned, we did not have the data to deliver the solutions.

LETS LOOK TO the future. There are two grounds for optimism.

The first is that data, always the lifeblood of AI, is now flowing. Kaggle, a machine-learning and data science platform is hosting the Covid-19 Open Research Dataset. CORD-19, as it is known, compiles relevant data and adds new research into one centralized hub. The new data set is machine readable, ng it easily parsed for AI machine learning purposes. As of publication, there are more 128,000 scholarly articles on Covid-19, coronavirus, SARS, MERS, and other relevant terms.

The second is that medical scientists and computer scientists across the world are now laser-focused on these problems. Peter Diamandis, founder of the XPrize Foundation, estimated that up to 200 million physicians, scientists, nurses, technologists, and engineers are now taking aim at Covid-19. They are running tens of thousands of experiments and sharing information “with a transparency and at speeds weve never seen before.”

The Covid-19 Research Challenge, also hosted on Kaggle, aims to provide a broad range of insights about the pandemic, including its natural history, transmission data and diagnostic criteria for the virus, and lessons from previous epidemiological studies to help global health organizations stay informed and make data-driven decisions. The challenge was released on March 16. Within five days it had already garnered more than 500,000 views and been downloaded more than 18,000 times.

In the first month of the outbreak in China, Alibaba released an AI algorithm trained on more than 5,000 confirmed coronavirus cases. Using CT scans, it can diagnose patients in 20 to 30 seconds. It can also analyze the scans of diagnosed patients and quickly assess health declines or progress, based on signs like white mass in the lungs. Alibaba opened its cloud-based AI platform to medical professionals around the world, working with local partners on anonymous data for deployment, including modules for epidemic prediction, CT Image analytics, and genome sequencing for coronavirus.

Deep learning—the capability to process massive, multi-model data at high speeds—presents one of the most far reaching opportunities for AI. Deep neural networks, a subtype of AI, have already been used to produce accurate and rapid algorithmic interpretation of medical scans, pathology slides, eye exams, and colonoscopies. I see a clear roadmap of how AI, accelerated by the pandemic, will be infused into health care.

THE POTENTIAL GOES beyond diagnosis and treatment.

Getting appointments, paying insurance bills, and other processes should be much less painful. AI combined with robotic process automation can analyze workflows and optimize processes to deliver significantly more efficient medical systems, improve hospital procedures, and streamline insurance fulfillment. To address the pandemic, AI could automate and accelerate pre-diagnostic inputs by crunching texts, languages, and numbers at machine-level quantity and precision.

With sufficient data as a foundation, AI can also establish health data benchmarks for individuals and for population. From there, its possible to detect variations from the baseline. That, in turn, positions us to identify potential pandemics early. Its not easy. Systems need to be connected so that early alert and response mechanisms can be truly effective. That appeared to be a shortcoming in the early days of the coronavirus outbreak.

There are already huge opportunities for using AI models and algorithms for new drug discovery and medical breakthroughs in genomic sequencing, stem cells, CRISPR, and more. In todays pharmaceutical world, there is a hefty price tag to developing a treatment. A huge part of this cost is eaten up by the money and time spent on unsuccessful trials. But with AI, scientists can use machine learning to model thousands of variables and how their compounded effect may influence the responses of human cells.

These technologies are already being used in the hunt for a Covid-19 vaccine and other therapies. Insilico Medicine, a Hong Kong-based AI company specializing in drug discovery, was among the first companies to react to Covid-19. The company used its generative chemistry AI platform to design new molecules to target the main viral protein responsible for replication. It published the molecules on February 5. AI and machine learning are ushering in an era of faster and cheaper cures for mankind. Drug discovery and the pharmaceutical industry as a whole will be revolutionised.

EARLY ONE WINTER morning in the year 2035,I wake up and notice a bit of a sore throat. I get up and walk to the bathroom. While I brush my teeth, an infrasensor in the bathroom mirror takes my temperature. A minute after I finish brushing my teeth, I receive an alert from my personal AI physician assistant showing some abnormal measurements from my saliva sample and that I am also running a low fever. The AI PA further suggests that I take a fingertip needle touch blood test. While the coffee is brewing, the PA returns with the analysis that I might be coming down with the flu, one of the two types around this season. My PA suggests two video calltime slots with my family doctor, should I feel the need to consult her. She will have all the details of my symptoms when I make the call. She prescribes adecongestant and paracetamol which is delivered to my door by drone.

That future is not as far off as it seems. Soon, as medical science and computer science further converge, we will move into an era of fully autonomous AI when we may expect people to choose wearables, biosensors, and smart home detectors to keep them safe and informed. And, as data quality and diversity increase from the wearables and other internet-of-things devices, a virtuous cycle of improvements will kick in.

In this world a novel coronavirus could be tracked, traced, intercepted, and cut off before it got going. In perhaps 15 years, many of us will have AI personal assistants in our households to keep us supported for our families day-to-day health issues. Robots or drones will deliver medication to our doors. If a surgery or some other medical intervention is needed, usually it will be a robot performing or assisting ahuman surgeon or doctor.

In this future doctors and nurses will focus more on the human tasks that no machine can do. The medical professionals or compassionate caregivers will combine the skills of a nurse, medical technician, social worker, and even psychologist. They will operate the AI-enhanced diagnostic tools and systems, but they will concentrate on communicating with patients, consoling them in times of trauma, and emotionally supporting them through their treatment.

In all this there are the key issues of privacy and data protection, particularly when it comes to patients’ records. It would be irresponsible to let useful data sit in their own isolated compartments, instead of extracting their usefulness to serve the progress of our societies. I am a big proponent of using innovative technological solutions to solve newly arisen technology issues, and the good news is that there has been progress made in federated learning, also known as distributed learning. In this framework, patients’ data is stored and never leaves their host health system or hospitals or personal devices, as machine learning models are trained from separate datasets, processed and combined subsequently. Technologies, such as federated learning, homomorphic encryption, and trusted hardware execution environments would further ensure data is computed, transmitted, and stored to meet preferred settings, as privacy requirements vary around different countries and cultures.

IF NOTHING ELSE,Covid-19 has proven that our shared challenges call for AI that recognizes how intertwined our destinies are. In the past global collaboration has led to the eradication of smallpox and the near-eradication of polio. As we work toward the goal of mitigating, treating, and eradicating the pandemic, it is clear that public health does not stop at national borders. Medicine is an arena where every country will benefit from building on, and with, others’ research. The whole world’s data will generate the most robust insights into health and disease.

AI will help ensure we will be better prepared for the next pandemic.It will need medical scientists, AI scientists, investors, and policy makers to collaborate. Venture capital is going to pour into healthcare and provide fresh impetus and focus for smart entrepreneurs and researchers. And, perhaps, as our brightest minds work on this challenge together, we can emerge acknowledging that our common enemy is not each other but a virus. It will take a planet to move our global healthcare systems to the next level.

本文相关词条概念解析:

医疗

医疗是指1、医治,2、疾病的治疗。中华医史几千年,而这个字眼是在近几十年才出现,其实这是为了与国际接轨而新生的字眼,之前大多使用治疗。然而医疗也包含保健内容。医疗事故是指医疗机构及其医务人员在医疗活动中,违反医疗卫生管理法律、行政法规、部门规章和诊疗护理规范、常规,过失就会造成患者人身损害的事故。从目前的调研信息反映来看,国内大部分医疗资源较为发达的地区已经开始了区域医疗卫生服务共同体的试点工作,目前此项国内还处于政府主导阶段,应此其组成部分主要为部属医院、市属医院、部队医院、社区卫生院和乡镇卫生院,民营医院和私人诊所暂时未加入其中。

网友评论

相关文章