Ensuring the patency of arteriovenous fistulas (AVFs) is critical for hemodialysis patients, yet traditional monitoring methods rely heavily on invasive procedures, leading to patient discomfort and increased healthcare resource consumption. A research team led by Dr. Wu Yen-Wen at Far Eastern Memorial Hospital has developed an innovative AI-assisted auscultation model, “A Novel Deep Learning Auscultation Model for Detecting Stenosis in Arteriovenous Fistula and Graft of Hemodialysis,” which utilizes deep learning and sound analysis to detect AVF stenosis in real time. This breakthrough technology enhances diagnostic accuracy, reduces the need for unnecessary ultrasound or catheter-based examinations, and improves patients’ quality of life while optimizing healthcare efficiency.
AVF occlusion can significantly impact dialysis efficiency and pose life-threatening risks. Current monitoring methods, such as ultrasound and catheter-based assessments, require skilled medical professionals and are not widely available at dialysis clinics, forcing patients to make frequent hospital visits. Additionally, traditional visual and auscultation-based assessments depend heavily on physician expertise and lack objective data, leading to inconsistencies in stenosis evaluation and treatment decision-making.
The AI model employs digital stethoscope technology to record 10-20 seconds of blood flow audio at the suspected stenosis site. Using Mel-Frequency Cepstral Coefficients (MFCCs) to extract 26-dimensional audio features, the system integrates Convolutional Neural Networks (CNN) and Deep Neural Networks (DNN) to accurately assess AVF stenosis severity. Clinical validation has demonstrated sensitivity and specificity rates of up to 94% and 86%, significantly outperforming traditional auscultation and existing monitoring techniques. The system’s non-invasive, highly accurate, and user-friendly nature makes it ideal for dialysis centers, hospitals, and home-based care. Furthermore, the technology can be integrated with mobile applications and remote monitoring platforms, enabling patients to continuously track AVF health, minimizing unnecessary clinic visits, and reducing overall healthcare expenditure.
Dr. Wu emphasized that this AI-powered auscultation model not only assists healthcare professionals in making more precise clinical decisions but also paves the way for remote AVF monitoring, alleviating the burden of frequent hospital visits. The research team is actively seeking commercial partnerships with insurance providers, health-tech companies, and medical institutions to develop a mobile application and cloud-based platform for comprehensive AVF management. This innovation aims to establish an end-to-end AVF care model, encompassing surgical intervention, AVF maturation, and long-term follow-up, ultimately ensuring sustained vascular access health for dialysis patients.
Resource:《新創動態》AI聽聲辨位檢測動靜脈廔管狹窄 居家洗腎更安全
