I-Corps: Automated sound analysis tool for early detection of arteriovenous fistula stenosis/failure in hemodialysis patients

这个I-Corps项目更广泛的影响/商业潜力是生物医学声学分析软件的开发。身体发出的声音往往反映了潜在的病理生理过程，容易收集但难以解释。提出的概念是彻底改变听诊这一体检的重要组成部分，将听诊从定性的艺术转变为定量的科学。潜在的应用包括分析心音、血流声、肺声、咳嗽声、肠声、神经传导声信号和肌肉细胞收缩声信号。从生物医学声学分析中及早发现潜在问题，可以启动必要的预防措施，减少下游成本，提高患者的生活质量。I-Corps项目基于人工智能（AI）系统的开发，该系统旨在通过生物医学声学分析检测血液透析患者早期动静脉（AV）瘘狭窄的迹象。在血管异常狭窄的地方，湍流产生不同于层流的频率、振动和声音特征。提出的人工智能软件系统包括一个机器学习模型，该模型可用于根据声学信号及其沿血管通道的相对变化将瘘管分类为专利或狭窄。机器学习模型是根据使用数字听诊器收集的房室瘘记录和双超声验证的血流状态进行训练的。在之前的研究中，机器学习模型在分类房瘘通畅方面的准确率达到80%。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this I-Corps project is the development of biomedical acoustic analysis software. Sounds produced by the body often reflect underlying pathophysiological processes and are easy to collect but difficult to interpret. The proposed concept is to revolutionize auscultation, an important element of the physical exam, and transform it from a qualitative art to a quantitative science. Potential applications include the analysis of heart sounds, blood flow sounds, lung sounds, cough sounds, bowel sounds, nerve conduction sound signals, and muscle cell contraction sound signals. Early detection of potential problems from biomedical acoustic analysis may initiate the necessary preventative actions, mitigating downstream costs and improving patient quality of life. The I-Corps project is based on the development of an artificial intelligence (AI) system that is designed to detect signs of early stenosis of arteriovenous (AV) fistulas in hemodialysis patients using biomedical acoustic analysis. At points of abnormal blood vessel narrowing, turbulent flow produces distinct frequencies, vibrations, and sound characteristics from laminar flow. The proposed AI software system comprises of a machine learning model that may be used to classify fistulas as either patent or stenotic based on the acoustic signals and their relative variations along the vascular access. The machine learning model is trained on AV fistula recordings collected using a digital stethoscope and flow statuses validated by duplex ultrasound. In previous studies, the machine learning model achieved an accuracy of 80% in classifying AV fistula patency.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.