Developing clinical decision support tools to characterize neurodegenerative disorders using biomedical speech signal processing and statistical machi

背景全球人口老龄化，提出了重要的社会挑战和国家卫生系统紧张，以满足日益增长的需求，医疗服务。包括智能手机和智能手表在内的新技术的兴起提供了一个独特的机会，通过收集额外的信号模式来彻底改变当代医疗保健服务，而无需人们频繁前往诊所。语音是一种易于收集的信号模态，需要最少的设备，并且已被证明可以传达关于神经退行性疾病（如帕金森病（PD））的临床重要信息。我们之前已经证明，语音可以用来区分健康对照与帕金森病患者，并复制标准症状严重程度度量通用帕金森病评定量表（Universal Parkinson's Disease Rating Scale，简称CAFRS）[1]。我们还证明了使用语音来评估远程PD康复[2]。此外，我们已经报告了对理解语音信号退化的发声生物力学的初步探索，从而获得了对PD症状严重程度进展的更机械的见解[3]。最近，我们已经表明，语音可以用作与遗传信息相关的PD的早期生物标志物[4]。总体而言，生物医学语音信号处理领域正在迅速扩大，并在过去几年中产生了相当大的研究兴趣。该项目的框架是进一步研究语音信号的潜力，目的是研究和监测神经退行性疾病的表现和进展，如帕金森病和阿尔茨海默病。学生将应用我们以前开发的算法，并扩展改进神经退行性疾病特征的方法。以前的研究只对比了一组神经退行性疾病和健康对照组;他们没有开发出鉴别诊断的工具（即解决区分疾病的问题并考虑潜在的共病）。该项目将扩展以前的工作，以研究不同神经退行性疾病对语音信号的影响，以提高对疾病进展和治疗计划的理解。我们拥有来自美国、澳大利亚和西班牙的临床合作者的丰富数据资源，这些数据很容易获得，并且之前已在我们小组的出版物中使用。此外，我们在西班牙马德里的临床同事（由共同主管维克托涅托-卢伊斯领导）已经开始收集一系列神经退行性疾病的数据，包括语音信号和疾病特异性临床标志物。被招募的学生将主要致力于开发新的时间序列，信号处理和模式识别算法，并扩展统计机器学习算法，以开发一个强大的用户友好的临床决策支持工具来表征神经退行性疾病。培训结果对临床实践和数据分析界面上的问题的实际理解，包括两端都有利基术语的语言障碍发展时间序列分析，信号处理和统计机器学习方面的专业知识，以解决大规模的挑战性问题编程技能：将算法概念转化为软件工具，并开发可供专家使用的界面

BackgroundThe population is aging globally, presenting important societal challenges and straining national health systems to meet increasing demand for healthcare delivery. The rise of new technologies, including smartphones and smartwatches, provides a unique opportunity to revolutionize contemporary healthcare delivery through the collection of additional signal modalities, without requiring frequent physical visits of people into clinics. Speech is a signal modality which is easy to collect, requires minimal equipment, and has been shown to convey clinically important information on neurodegenerative conditions such as Parkinson's Disease (PD). We have previously shown that speech can be used to differentiate healthy controls from people with Parkinson's disease and replicate the standard symptom severity metric Universal Parkinson's Disease Rating Scale (UPDRS) [1]. We have also demonstrated the use of speech to assess remote PD rehabilitation [2]. Furthermore, we have reported initial explorations towards understanding phonation biomechanics of speech signal degradation, thus gaining a more mechanistic insight into PD symptom severity progression [3]. More recently, we have shown that speech could be used as an early biomarker of PD associated with genetic information [4]. Overall, the field of biomedical speech signal processing is rapidly expanding and has generated considerable research interest over the past few years. Aims The framework of this project is to further investigate the potential of speech signals, with the goal of studying and monitoring the manifestation and progression of neurodegenerative diseases, such as Parkinson's disease and Alzheimer's disease. The student will apply algorithms we have previously developed and extend approaches towards improving the characterization of neurodegenerative diseases. Previous studies have only contrasted a group with a neurodegenerative disorder and healthy controls; they have not developed tools towards differential diagnosis (i.e. tackling the problem of differentiating diseases and considering potential co-morbidities). This project will extend previous work to investigate imprints of different neurodegenerative disorders on speech signals towards improving understanding of disease progression and treatment planning. We have rich data resources from clinical collaborators based in the US, Australia, and Spain, which is readily available, and which has been previously used in publications in our group. Moreover, our clinical colleagues in Madrid, Spain (led by co-supervisor Victor Nieto-Lluis) have already started data collection across a range of neurodegenerative disorders, including speech signals and disease-specific clinical markers. The recruited student will be primarily working on developing novel time-series, signal processing, and pattern recognition algorithms, and extending statistical machine learning algorithms to develop a robust user-friendly clinical decision support tool to characterize neurodegenerative disorders.Training outcomes Practical understanding of the problems at the interface of clinical practice and data analytics, including the language barrier with niche terminology on both ends Developing expertise in time-series analysis, signal processing, and statistical machine learning to tackle large-scale challenging problems Programming skills: transforming algorithmic concepts to software tools, and developing interfaces which can be used by experts