Research Project
10255460
Project Summary/Abstract This SBIR Phase I project will develop a deep learning-based algorithm to analyze the sound of blood flow in mature arterio-venous (AV) fistulas used for hemodialysis access. This monitoring tool can help to identify fistulas with impending failure in patients who are in need of surgical intervention to ensure patency of the patient?s hemodialysis access. The specific aims of the study are (1) to create the world?s first deep learning-scale database of mature AV fistula sounds paired with ultrasound imaging from hemodialysis patients, and (2) develop and evaluate the performance of a deep learning classification model trained via semi-supervised learning to discriminate between patients with patent fistulas and patients with failing fistulas. By integrating this deep learning algorithm into Eko?s mobile and cloud software platform, we anticipate this algorithm will enable better monitoring for failing fistulas. During Phase I of the project we will recruit study subjects in vascular surgery and interventional radiology clinics at Columbia University Medical Center. The follow-up SBIR Phase II project will extend our algorithm to account for (1) broader forms of hemodialysis access including AV grafts, which often fail more frequently than fistulas, and premature fistulas, which are often abandoned, and (2) longitudinal collection and analysis of hemodialysis access sounds at commercial dialysis centers and in patients? homes.
