Novel data analytics tools combined with high-resolution cervical auscultation are needed to instrumentally screen for dysphagia

吞咽困难（吞咽障碍）是与衰老和神经系统疾病相关的最常见问题，影响着数百万加拿大人的日常生活。吞咽困难的风险目前是通过筛查来评估的，在诊断金标准的视频透视检查之前，但许多无声吸气的患者通过了最初的筛查。PI的长期研究目标是利用计算方法和仪器，并将创新的工程研究转化为吞咽困难的临床解决方案。因此，本提案的短期研究目标是将全新数据分析工具的进展与高分辨率宫颈听诊（HRCA -加速度计和颈部麦克风记录）相结合，以辅助筛查吞咽困难。教育目标是通过结合信号处理，机器学习和仪器仪表，为高素质人才（HQP）创造跨学科的培训机会。HQP还将获得与老龄化和神经系统疾病相关的主要健康问题的广泛知识，这些问题是加拿大医疗保健支出的主要来源。该计划是对当前信号处理工作的重大背离，重点关注吞咽困难的信号处理和机器学习方法的创新。我们的变革性方法解决了当前信号处理的障碍：首先，传统方法无法将HRCA信号分析结果转化为吞咽障碍的有效临床测量。为了解决这一重大问题，我们将基于卷积和递归神经网络以及HRCA信号的时频表示来创新新的深度学习方法。其次，传统的信号处理方法无法对吞咽函数进行逆建模。为了解决这个问题，我们的新方法是基于应用于信号的生成对抗网络。拟议研究的关键变革方面包括吞咽困难数据分析工具的基础理论进步，同时将研究转化为临床应用工具，同时也适用于其他工程（例如人工智能）和临床领域（例如电生理学）的工具。该提案支持NSERC 2020战略计划，通过开发数据驱动的医疗保健方法和跨学科培训机会，重点培养HQP的创造力、转化和沟通技能。此外，PI还将通过扩大多伦多大学现有的机会，积极招募和鼓励妇女和代表性不足的少数族裔HQP参与拟议的项目。最后，PI与临床合作伙伴的密切合作提供了通过研究出版物、教程、研讨会、临床大查班和技术转让活动将学术成果转化为主流临床实践的机会。

Dysphagia (swallowing disorders), the most common issue associated with aging and neurological disorders, affects the daily lives of millions of Canadians. Dysphagia risk is currently assessed via screening, before the diagnostic gold-standard videofluoroscopic test, but many patients who silently aspirate pass initial screens. The PI's long-term research goal is to utilize computational approaches and instrumentation and translate innovative engineering research to clinical solutions for dysphagia. Therefore, the short-term research goal of this proposal is to combine the advances in fundamentally new data analytics tools with high-resolution cervical auscultation (HRCA - accelerometer and microphone recordings from the neck) to instrumentally screen for dysphagia. The educational objective is to create interdisciplinary training opportunities for highly qualified personnel (HQP) by combining signal processing, machine learning, and instrumentation. HQP will also acquire extensive knowledge of major health issues associated with aging and neurological disorders, which are the major contributors to healthcare expenditures in Canada. The proposed program is a major departure from the current signal processing efforts by focusing on the innovation of signal processing and machine learning approaches for dysphagia. Our transformative approach addresses current signal processing obstacles: First, classical approaches cannot translate HRCA signal analysis results to a validated clinical measure of swallowing impairment. To resolve this major issue, we will innovate new deep learning approaches based on convolutional and recursive neural networks along with time-frequency representations of HRCA signals. Second, inverse modeling of the swallowing function is unfeasible with traditional signal processing methods. To address this issue, our new approach is based on generative adversarial networks applied to signals. The key transformative aspect of the proposed research consists of fundamental theoretical advancements of data analytics tools for swallowing difficulties, while translating the research into clinically applicable tools, but also tools applicable in other engineering (e.g., artificial intelligence) and clinical fields (e.g., electrophysiology). This proposal supports the NSERC 2020 strategic plan by developing data-driven healthcare approaches and interdisciplinary training opportunities focused on the cultivation of HQP's creativity, translational, and communication skills. Furthermore, the PI will also actively recruit and encourage participation of women and underrepresented minority HQP to the proposed project by expanding currently available opportunities at the University of Toronto. Lastly, the PI's close collaboration with clinical partners offers the opportunity to transfer academic results into mainstream clinical practices via research publications, tutorials, workshops, clinical grand rounds, and technology transfer activities.