中国首个坚持原创严肃内容的健康医疗信息化新媒体
最新消息:欢迎您,亲爱的读者!您可以通过QQ号或新浪、腾信微博账号直接在评论处登录,发表评论并选择转发到微博、QQ空间。

张家杰教授:数据科学 = 信息学

专栏 HIT金子 1618浏览 评论

来源:HIT专家网    作者:德州大学休斯顿健康科学中心生物医学信息学院院长、教授  张家杰

张家杰博士

信息学与数据科学之间的关系是什么?信息学与数据科学之间的关系其实非常简单,它们是1:1的等价关系:信息学= 数据科学。

在医疗行业,生物医学信息学就是加入生物医学和医疗卫生成分的数据科学。数据科学是一个最近出现在大数据时代的新名词,而生物医学信息学,即生物医学和医疗卫生领域的数据科学,已经有几十年的历史了。

生物医学信息学是一个高度跨学科的领域,涉及以下学科:

  • 临床科学与实践:医学、护理、牙科、药学、人口健康
  • 公共卫生和社区健康
  • 计算机科学与工程:数据库、算法、编程、人工智能、机器学习(包括深度学习)、神经网络、认知计算、分布式计算、云计算、自然语言处理和文本处理、安全性、可视化、移动设备、远程医疗、物联网等
  • 认知科学
  • 数学和生物统计学
  • 社会和行为科学
  • 管理科学
  • 健康信息技术政策和法律问题

生物医学信息学强调一些关键的过程(process),包括:采集、存储、通信、处理、集成、分析、挖掘、检索、解释和演示。这些过程将数据转化为信息,再从信息到知识,从知识到智能。数据、信息、知识和智能被称为实体(entity)。

有了这些实体后,就可以执行描述性(descriptive)、预测性(predictive)和指令性(prescriptive)的任务或功能(function)。在一般的数据科学或信息学中,这些过程的实体和功能可以应用于任何领域。对于生物医学信息学来说,这些应用领域是生物医学发现、医疗保健和疾病预防。

SBMI的教师,学生和研究人员专注于这些过程、实体和功能的创新, 正在生物医药和医疗卫生领域里进行深入的数据科学的研究。

我们来仔细看看这个框架是如何工作的:

信息学=数据科学

  • 数据(data)是未经翻译的、未经处理的、没有意义的原始符号、信号或像素。例如,“101”可能有几件含义:十进制数一百零一,二进制数五,三个像素的值,甚至高速公路的标签。没有上下文及周边环境,大多数数据是没有意义的。
  • 信息(information)是解释过的、有意义的数据。例如,一旦我们知道101的度量是华氏度,我们立即将这个数字与温度相关联。信息可以提供一个描述性(descriptive)功能,告诉你发生了什么,在什么时候,和谁有关系。
  • 知识(knowledge)是经过验证的、有组织的信息。假设我们知道101°F是成人的口腔体温,我们立即会知道这表明身体状态异常(发烧),这种关系是在医学实践和研究中得到验证的知识。知识可以提供预测性(predictive)功能,它可以告诉您可能会发生什么。比如说101°F华氏度的成人口腔体温(38.3°C)预示着身体状况不正常,有低烧。
  • 智能(intelligence)是可操作的知识。体温为101°F的成年人应服用退烧药物,进行进一步的评估和诊断,如果不是简单的感冒,可能需要去看医生。智能可以提供指令性(prescriptive)的功能与建议,告诉你需要做什么。101°F成人口腔体温告诉您需要给患者退烧药或进一步诊断。

【作者简介】

 张家杰教授:德州大学休斯顿健康科学中心“生物医学信息学院”院长,教授,Glassell 家族基金会Distinguished Chair, 美国国家认知信息学和决策中心(National Center for CognitiveInformatics and Decision Making)主任, 美国医学信息学院(American College of MedicalInformatics)Fellow,中国科技大学少年班毕业,加州大学圣地亚哥分校认知科学博士。

【英文原文】

When examining the connection between informatics and data science, the ratio is rather simple– 1:1.Informatics is the equivalent of data science.

Biomedical informatics is an amalgam of data science with both biomedicine and health components added in. Data science is a recent name that grew out of the emergence of big data, although biomedical informatics, i.e., data science in biomedicine and healthcare, has been around for several decades. The field of biomedical informatics is also an interdisciplinary field that involves:

·       Clinical science and practice: medicine, nursing, dentistry, pharmacy, population health

·       Public and community health

·       Computer science and engineering: database, algorithm, programming, artificial intelligence, machine learning (including deep learning), neural network, cognitive computing, distributed computing, cloud computing, natural language processing and text processing, security, visualization, mobile devices, sensors, internet of things, etc.

·       Cognitive science

·       Mathematics and biostatistics

·       Social and behavioral sciences

·       Management science

·       Health information technology policy and legal issues

Within biomedical informatics, there is an emphasis on certain key processes; acquisition, storage, communication, processing, integration, analysis, mining, retrieval, interpretation, and presentation. These processes transform data to information to knowledge to intelligence; these are entities.

Once researchers have entities for evaluation, the next step is to perform descriptive, predictive, and prescriptive tasks or functions. In general data science or informatics, these processes, entities, and functions can be applied to any domains. For biomedical informatics, the application domains are biomedical discovery, healthcare delivery, and disease prevention.

Focusing on innovations in these processes, entities, and functions, faculty, students, and researchers at SBMI are performing in-depth research studies within the field of data science in biomedicine and healthcare.

Let’s take a closer look at how this framework works:

  • Data is un-interpreted, unprocessed, meaningless raw symbols, signals or pixels. For example, “101” could mean several things; the decimal number one hundred and one, the binary number five, the values of three pixels or even a label for a highway. Without context, most data types are meaningless.
  • Information is interpreted data or data with meaning. For example, once we know that the metric for 101 is degrees in Fahrenheit, we immediate correlate the number to temperature. Information provides a descriptive function and tells you what happened, at what juncture and for whom.
  • Knowledge is organized information that is justified or validated. Let’s say that we also know that 101 °F is an adult oral temperature. Immediately, we know that this indicates an abnormal body status (fever) and this relation is validated in medical practice and research. Knowledge provides a predictive function and tells you what might happen. An adult oral temperature of 101 °F predicts that the body status is irregular.
  • Intelligence is actionable knowledge. An adult with a 101°F temperature should take fever medication, have further assessment and diagnosis performed and may need to see a doctor if it is not a simple cold. Intelligence provides prescriptive function and tells you what needs to be done. 101 °F adult oral temperature prescribes the action of taking fever medicine and further assessment.

【责任编辑:谭啸】

 

您必须 登录 才能发表评论!