残疾科学家_数据科学与残疾：通过创新加强护理

残疾科学家

Could the time it takes for you to water your houseplants say something about your health? Or might the amount you’re moving around your neighborhood reflect your mental health status?

您给植物浇水所需的时间能否说明您的健康状况？ 还是您在附近移动的金额反映出您的心理健康状况？

When translated into data and analyzed, these measures could indeed provide insights into the health of people living with chronic physical or mental illness or disabilities.

当转换为数据并进行分析时，这些措施的确可以提供对患有慢性身体或精神疾病或残疾的人的健康的见解。

Data science has presented new possibilities for greater independence, improved care and better outcomes. Whether it’s healthcare organizations using data to deliver care more effectively, developers creating data-driven apps that help identify early warning signs of illness, or inventors creating new sensors and devices to detect health issues, all these tools rely on data science techniques to help people living with disabilities or mental illness. Here are some examples of this kind of innovation that I found especially intriguing.

数据科学为增强独立性，改善护理和改善结局提供了新的可能性。 无论是医疗机构使用数据来更有效地提供护理，开发人员创建数据驱动的应用程序以帮助识别疾病的早期征兆，还是发明家创建新的传感器和设备来检测健康问题，所有这些工具都依赖于数据科学技术来帮助人们患有残疾或精神疾病的人。 以下是一些我特别感兴趣的创新示例。

在家中异常检测 (Anomaly Detection at Home)

People who are able to live mostly independently could benefit from advances in smart home technologies and sensors embedded in their living environments. Anomaly detection is a key data science technique used in many different contexts, such as detecting credit card fraud and monitoring manufacturing processes. Researchers are also working on using anomaly detection to identify deviations from typical behavior for people living in “smart” homes equipped with sensors, such as cameras and computerized pill dispensers that report on medication habits.

能够独立生活的人们可以从智能家居技术和生活环境中嵌入的传感器中受益。 异常检测是在许多不同情况下使用的关键数据科学技术，例如检测信用卡欺诈和监视制造过程。 研究人员还致力于使用异常检测来识别居住在配备传感器的“智能”房屋中的人们与典型行为的偏差，例如摄像机和报告用药习惯的计算机化药丸分配器。

In one study, researchers established a baseline for movement and behavior at home by having volunteers complete “Instrumental Activities of Daily Living,” or routine home activities like watering plants or microwaving food. The researchers constructed an activity graph representing how that behavior looked in time and physical space for each volunteer.

在一项研究中 ，研究人员通过让志愿者完成“日常生活中的工具性活动”或日常的家庭活动(例如浇水或微波烹饪食物)来建立家庭活动和行为的基准。 研究人员构建了一个活动图，表示每个志愿者在时间和身体空间上如何看待这种行为。

Knowing that baseline, the researchers’ anomaly detection algorithms could identify deviations that could represent cognitive or physical concerns. These researchers plan to integrate this capability into a real-time monitoring application that could show both sudden or longer-term deviations from a person’s baseline behavior patterns at home.

知道了基线之后，研究人员的异常检测算法可以识别出可能代表认知或生理问题的偏差。 这些研究人员计划将该功能集成到实时监视应用程序中，该应用程序可以显示与一个人的基准行为模式在家中突然或长期的偏差。

The anomalous pattern on the bottom shows that the resident is taking an unusually long time to fill a watering can in order to complete the task of watering plants. That anomaly could reveal larger challenges in completing household activities. Image from “Anomaly Detection of Elderly Patient Activities in Smart Homes using a Graph-Based Approach,” by Ramesh Paudel, William Eberle, and Lawrence B. Holder

底部的异常模式表明，该居民花了非常长的时间来填充喷壶，才能完成给植物浇水的任务。 这种异常可能揭示完成家庭活动中的更大挑战。 图片来自 “ Ramesh Paudel，William Eberle和Lawrence B. Holder的 “使用基于图的方法对智能家居中的老年患者活动进行异常检测”的 图像

帕金森氏病的聚类分析 (Cluster Analysis for Parkinson’s Disease)

It’s hard for people with Parkinson’s disease and their caregivers to identify, track and respond to the many different ways the disease can manifest and progress. Wearable sensors and the data they generate could help address this challenge, however.

对于帕金森氏病及其护理人员来说，很难识别，跟踪和应对这种疾病表现和发展的许多不同方式。 但是，可穿戴式传感器及其生成的数据可以帮助应对这一挑战。

For example, this research study explained how a wearable device could gather data for patients and clinicians, including tracking of activity, sleeping, falls, gait characteristics and “freezing” (a temporary inability to move that is experienced by those with Parkinson’s). That data could offer more comprehensive insights than those available in a brief office visit with a medical provider.

例如，这项研究解释了可穿戴设备如何收集患者和临床医生的数据，包括跟踪活动，睡眠，跌倒，步态特征和“冻结”(帕金森氏症患者暂时无法移动)。 与在与医疗服务人员的简短办公室访问中所获得的见解相比，这些数据可以提供更全面的见解。

However, the researchers also point out that better analytic techniques are needed to make full use of this kind of data. They highlight the potential for unsupervised clustering techniques that could make sense of the large quantity of data gathered by a constantly worn device. Their research used t-Distributed Stochastic Neighbor Embedding (t-SNE), which is a way to visualize high-dimensional datasets in two dimensions. The image below shows an example.

但是，研究人员还指出，需要更好的分析技术来充分利用此类数据。 他们强调了无监督群集技术的潜力，这些技术可以理解不断磨损的设备收集的大量数据。 他们的研究使用t-分布随机邻居嵌入 (t-SNE)，这是一种可视化二维高维数据集的方法。 下图显示了一个示例。

“A Perspective on Wearable Sensor Measurements and Data Science for Parkinson’s Disease,” by Ricardo Matias, Vitor Paixão, Raquel Bouça, and Joaquim J. Ferreira “帕金森氏病可穿戴式传感器测量和数据科学的视角”

The image shows the clustering of movement data from a wearable sensor worn by a patient with Parkinson’s. The red Xs represent movement during an “off state” (the term used to describe a period when a patient doesn’t respond well to levodopa medication); the blue dots represent movement during an “on state,” when medication was working well. The clustering approach shows the difference between those times and could help identify patterns in a patient’s movement that would be useful in refining a personalized treatment strategy.

该图像显示了帕金森氏症患者佩戴的可穿戴传感器的运动数据聚类。 红色的X代表“关闭状态”(该状态用于描述患者对左旋多巴药物React不良的时期)的运动； 蓝点表示药物正常运作时“开启状态”下的运动。 聚类方法显示了这些时间之间的时差，可以帮助确定患者运动的模式，这将有助于完善个性化治疗策略。

精神健康的地理位置 (Geolocation for Mental Health)

Do your movements in your local area reflect your mental health status? What about the number of phone calls you make or text messages you send?

您在当地的活动是否反映出您的心理健康状况？ 您拨打的电话或发送的短信数目如何？

Researchers have been studying the potential of smartphone data for treating mental illness for some time. Mobility, social interaction and survey responses could all be useful in helping people with mental illness manage and assist caregivers in knowing when an intervention might be timely.

一段时间以来，研究人员一直在研究智能手机数据在治疗精神疾病方面的潜力。 流动性，社交互动和调查回复都可以帮助精神疾病患者管理和协助看护者了解何时应该及时进行干预。

One pilot study of an app designed for people with schizophrenia used both active data collection (asking users about their symptoms) and passive data (e.g., geolocation, communication frequency). The data were integrated into a multivariate time series model and could be monitored for days when multiple features were “simultaneously and sufficiently anomalous.” Those occurrences could predict relapse and hospitalization for app users, potentially providing “early warning signs” for caregivers.

一个试验性研究设计用于精神分裂症患者一个应用程序的使用主动收集数据(要求用户对他们的症状)和无源数据(例如，地理位置，通信频率)。 数据被集成到一个多元时间序列模型中，并且可以在多个特征“同时且充分地异常”的情况下监控几天。 这些事件可能会预测应用程序用户复发和住院，从而可能为护理人员提供“早期预警信号”。

The image above shows the p values for different data points for one patient, collected in the study of an app for people with schizophrenia. There’s increased variability in the patient’s daily mobility, sociability and self-reported symptoms right before a relapse and hospitalization. Image from “Relapse prediction in schizophrenia through digital phenotyping: a pilot study,” by Ian Barnett, John Torous, Patrick Staples, Luis Sandoval, Matcheri Keshavan and Jukka-Pekka Onnela

上图显示了在针对精神分裂症患者的应用程序研究中收集的一名患者不同数据点的p值。 在复发和住院之前，患者的日常活动能力，社交能力和自我报告症状的变异性增加。 从图像 “通过数字表型精神分裂症复发预测的初步研究”，由Ian巴尼特，约翰Torous，帕特里克订书钉，路易斯桑多瓦尔，Matcheri Keshavan和的Jukka-佩卡Onnela

The researchers noted that mental illness is extremely varied and complex. They suggest that data-driven approaches may be best used in creating a personalized model for relapse, reflecting one individual’s unique patterns, versus trying to create a generalizable model that applies to everyone contending with the same disease.

研究人员指出，精神疾病极为多样化和复杂。 他们建议，以数据驱动的方法最好用于创建个性化的复发模型，以反映一个人的独特模式，而不是尝试创建适用于所有患有相同疾病的人的通用模型。

重要注意事项 (Important Considerations)

If you’ve been reading this and thinking, “But … privacy!” — yes, indeed. There’s certainly a great deal of personal information that is shared to use smart home sensors, wearable devices or smartphone apps as indicators of physical and/or mental health.

如果您一直在阅读此书并在思考，“但是……隐私！” - 确实是的。 当然，可以共享大量的个人信息以使用智能家居传感器，可穿戴设备或智能手机应用程序来指示身体和/或心理健康。

One study showed that older adults would be willing to share this kind of data with their primary medical professionals and “most trusted people,” and would be OK with the data being stored short term (defined in the study as up to one week). With regard to mental health, a worrisome 2017 study showed that fewer than half of the 116 available mental health smartphone apps had a privacy policy. A 2018 survey of people with a history of mental illness showed they had strong concerns about the external monitoring of their health, the potential for undesired sharing to social networks, and the risk of exposure of sensitive information to third parties or hackers.

一项研究表明，老年人愿意与他们的主要医疗专业人员和“最受信任的人”共享此类数据，并且可以将数据短期存储(在研究中定义为长达一周)。 关于心理健康，一项令人担忧的2017年研究表明，在116种可用的心理健康智能手机应用程序中，只有不到一半具有隐私政策。 2018年对有精神病史的人的调查显示，他们对外部监控自己的健康状况，社交网络中不希望有的共享的可能性以及敏感信息暴露给第三方或黑客的风险感到非常担忧。

Data science techniques will undoubtedly help address many health challenges, but the public may need a little reassurance about how their data will be used in this effort. Still, the potential for both personalized treatment and widespread insights into a wide variety of diseases is clear and exciting.

数据科学技术无疑将帮助解决许多健康挑战，但是公众可能需要在此方面将如何使用其数据方面有所放心。 尽管如此，个性化治疗和对多种疾病的广泛见解的潜力是显而易见而令人兴奋的。

This article was originally published on the Alteryx Community.

本文最初在 Alteryx社区上发布 。

翻译自: https://medium.com/swlh/data-science-and-disability-enhancing-care-with-innovation-35577b3c992a

残疾科学家