Exploring the Potential of Sound Analysis to Detect Stenosis in Arteriovenous Fistulae for Hemodialysis

Objectives: Native arteriovenous fistula (AVF) is the preferred vascular access for hemodialysis (HD), but it still fails in more than 40% cases within the first year. Stenosis is the main cause of AVF failure, but its detection and prediction are still open clinical challenges. The purpose of our study was to explore the potential of using sound analysis to reveal unique characteristics in AVF sounds and to detect potentially relevant changes over time. Methods: We acquired the sounds of 10 AVFs, 7 referred by the nephrologists as well-functioning and 3 as stenotic AVFs. We also acquired the sound of 1 AVF characterized by severe stenosis and we repeated the recording after surgical revision. Sounds were recorded using the Littmann Electronic Stethoscope 3200, using wide mode. Sounds were transmitted via Bluetooth to a laptop and frequency spectra were obtained using an in-house Matlab code embedding the Fast Fourier Transform and a high-pass filter with a cutoff frequency of 50 Hz, aimed at attenuating stethoscope’s noise and heart sounds’ frequencies. Results: In well-functioning AVFs we consistently found a low-frequency peak, located in the bandwidth 100-200 Hz, while no relevant spectral features were present at higher frequencies. On the other hand, stenotic AVFs showed high-frequency peaks, located in the bandwidth of 500-600 Hz. In the AVF with severe stenosis we found high-frequency peaks in the bandwidth 700-750 Hz, which were replaced by low-frequency peaks after AVF revision. Discussion: Sound analysis revealed unique characteristics in the frequency spectra of AVFs, allowing an objective discrimination between well-functioning and stenotic AVFs. This technique may bring the advantage of limiting clinician’s skill-dependency and personal interpretation. Despite being preliminary, our results suggest that sound analysis may be used to detect stenosis development and allow patients’ self-monitoring, resulting in early prediction of AVF failure.