EAGER: SCH: AI in Sleep for Rural and Aging Communities

尽管人工智能(AI)和机器学习已经变得司空见惯，但弱势群体往往没有投入到改善社区福祉和安全的关键人工智能工具的设计、开发和应用中。人工智能工具数据输入的多样性减少，削弱了技术和由此产生的输出的整体能力。该项目利用人工智能技术将患者与睡眠研究专业人员和专家连接起来，纵向获取睡眠健康信息。这一方法为无法定期咨询睡眠专家的人群提供支持，特别是农村和难以接触到的人群，如轮班工人和老年人。该项目还向人工智能工具提供独特的患者数据，以增加参与优化这项技术的患者群体，这将缓解人工智能工具的输入偏差。这一急切的研究努力为基于人工智能的医疗保健技术提供了一个独特的机会，因为它增加了人口投入的多样性，并改善了农村和需要持续求助睡眠专家的弱势人群的医疗保健机会。来自这些群体的个人将被招募，为该技术的整体输入数据做出贡献，产生广泛适用的输出结果。该项目使用了一种新合成和验证的人工智能技术，该技术使用睡眠阶段生理特征的频率作为输入，并分析结果输出的准确性。该项目的目标是：1)通过扩大收集的输入数据来优化人工智能技术的数据处理和算法分析，提高深度学习算法输出对公众的敏感度、精确度和适用性；2)通过将收集的信息与与农村和其他参与群体具有相似背景(例如年龄、性别和种族)的参与者的信息进行比较，确定人工智能技术的数据处理和解释的准确性以及独特人群的适应性；以及3)通过成功地适应专门的人工智能技术来增加获得所需人口服务的机会。该项目将探索在深度学习算法中使用人工神经网络来解释参与者在睡眠时收集的原始数据的有效性。人工智能工具产生的输出和睡眠研究(多导睡眠图)的评分将进行比较分析。PI将与睡眠专家合作，其中一些专家距离参与者的家或工作地点超过四个小时，以进行基线多导睡眠监测评分和原始数据分析。产出将与类似群体的队列数据进行比较。该项目的成果包括为分析中更可预测和更准确的人工智能分析输出功能提供更广泛的输入。此外，更好地获得专家护理和使用精确的分析工具将改善研究小组和普通民众的健康和福祉。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Although artificial intelligence (AI) and machine learning have become commonplace, vulnerable populations often do not have input to the design, development, and application of key AI tools that improve the well-being and safety of communities. Reduced variety in the data input of AI tools diminishes the overall power of the technologies and the resulting output. This project uses AI technology to connect patients with sleep research professionals and specialists to obtain sleep health information in a longitudinal fashion. This approach provides support to populations who do not have regular access to sleep specialists, particularly rural and hard to reach populations such as shift workers and aging adults. The project also provides unique patient data to the AI tools to increase the patient population groups involved in optimizing this technology, which will mitigate input bias of the AI tools. This EAGER research effort is a unique opportunity for AI-based technologies for healthcare by having additional variation in population input, and by improving healthcare access to rural and vulnerable populations with a need for sustained access to sleep specialists. Individuals from these groups will be recruited, contributing to the technology’s overall input data producing broadly applicable output results.This project uses a newly synthesized and verified AI technology that uses the frequencies of physiological characteristics during sleep stages as inputs and analyzes resulting outputs for accuracy. The goals of the project are to: 1) optimize the data processing and algorithmic analysis of AI technologies by broadening the input data collected, improving the sensitivity, precision, and applicability of the output from deep learning algorithms for the public, 2) determine the accuracy of the AI technology’s data processing and interpretation and adaptability from unique population groups by comparing collected information with that of participants with similar backgrounds (for example, age, sex, and race) as the rural and other participant groups, and 3) increase access to needed population services through successful adaptability of specialized AI technologies. The project will explore the effectiveness using artificial neural networks in deep learning algorithms to interpret the raw data collected from participants as they sleep. Output produced from the AI tool and scoring of sleep studies (polysomnography) will be comparatively analyzed. The PIs will collaborate with sleep specialists, some who are located more than four hours from the participants’ home or work locations, for baseline polysomnography scoring and raw data analysis. Outputs will be compared with cohort data from similar groups. The outcomes of this project include providing more broad-based input for an AI analysis output function that is more predictable and accurate in analysis. Further, improvement in the health and well-being of the study group and the general population will result from better access to specialist care and use of precision tools for analysis.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.