CI-New: An Open Speech Data Repository for Medical Prediction and Assessment of Neurological Disorders

这个社区研究基础设施项目将建立一个语音数据和语音处理算法的存储库，以支持旨在通过分析语音来检测神经疾病的研究。现有的检测神经疾病的技术成本很高，或者很难用于初级医疗服务。然而，神经疾病往往会在语音和语音产生中留下指纹，这表明语音信号分析可以提供临床信息，以预测某些疾病，诊断疾病，评估疾病进展或治疗方案的有效性。基于语音的评估将具有显著的优势，包括成本低、侵入性最小和易于使用。支持旨在更好地了解神经疾病和言语产生之间的联系的研究，以支持新的诊断工具可以带来巨大的好处。例如，了解帕金森氏病、小脑脱髓鞘和中风等疾病引起的语音和声音变化，可能为早期发现这些疾病的发病、进展和严重程度提供信息。然而，由于三个密切相关的问题，此类诊断工具仍然遥不可及：(1)缺乏对神经疾病和发声之间的关系的深入了解；(2)现有研究中使用的语音样本集较小且不完整；(3)缺乏能够进行神经评估的语音捕获、处理和分析工具和算法。该项目将提供基础设施来解决最后两个问题，以使研究能够解决第一个问题。该项目包括(1)设计、推出和管理语音样本库；(2)使用移动和基于网络的众包方法从所有人口部分的受试者收集大量语音样本和医疗数据；以及(3)设计、实施、评估和分发一套分析工具，为如何分析和解释语音提供统一的方法。与仅基于少数受试者或不包括来自不同年龄、性别或合并疾病的受试者的现有研究相比，由此产生的研究结果将是一个重大进步。储存库和附带的处理工具将使发声领域能够研究神经状况对言语的影响，从而改进诊断和评估工具。私人投资机构还将依靠拟议的工作来扩大其外联和教育工作，包括指导少数族裔和高中生、监督本科生研究人员以及根据拟议的资料库设计当前的课程。

This community research-infrastructure project will establish a repository of speech data and voice-processing algorithms to support research aimed at detecting neurological disorders by analyzing speech. Existing techniques to detect neurological disorders are costly or difficult to use in primary medical services. However, neurological disorders often leave a fingerprint in voice and speech production, suggesting that speech signal analysis could provide clinical information to predict certain diseases, diagnose illnesses and assess disease progression or the effectiveness of treatment regimens. Voice-based assessment would have significant advantages including low cost, minimal intrusiveness and ease-of-use. Supporting research aimed at better understanding the link between neurological conditions and speech production to enable new diagnostic tools can have immense benefits. For example, understanding the changes in speech and voice caused by diseases such as Parkinson's disease, cerebellar demyelination and stroke, may provide information for early detection of onset, progression and severity of these diseases. However, such diagnostic tools are still out of reach because of three closely related problems: (1) the lack of an in-depth understanding of the relationship between neurological disorders and phonation and (2) the small and incomplete sets of voice samples used in existing studies, and (3) the lack of voice capture, processing, and analysis tools and algorithms that enable neurological assessment. This project will provide infrastructure to address the final two problems in order to enable research to address the first.The project includes (1) design, rollout, and management of a repository of voice samples; (2) collection of a large number voice samples and medical data from subjects across all parts of the population using a mobile- and web-based crowdsourcing approach, and (3) design, implementation, evaluation and distribution of a set of analytical tools that will provide a unified approach to how speech is analyzed and interpreted. The resulting collection will be a major advance over existing studies that are based on only a few subjects or that do not include subjects from a variety of ages, genders or co-morbidities. The repository and accompanying processing tools will enable research in the area of phonation on the impact of neurological conditions on speech, leading to improved diagnostic and assessment tools. The PIs will also rely on the proposed work for extending their outreach and educational efforts, including mentorship of minority and high-school students, supervision of undergraduate researchers, and the design of current courses based on the proposed repository.