Electronic Stethoscope and Computer-aided Diagnostic System in Detecting Arteriovenous Fistula (AVF) Stenosis

Hemodialysis is a well-established method for removal of fluid and uremic toxins and prolonging survival in patients with end-stage renal disease (ESRD). The most common permanent vascular assesses are autogenous and non-autogenous arteriovenous (AV) fistula, referred as AV fistula and AV graft, respectively. In Taiwan, both incidence and prevalence of ESRD remain highest worldwide. However, frequent vascular stenosis/ thrombosis events occurred because of repeated needle punctures. These will exert a large economic burden to our society and psychological impacts on ESRD patients [1]. Therefore, it is essential to develop an easy and patient-centered method to monitor the flow status in either AV fistula or AV graft.

Physical examination and auscultation are simple monitoring methods to assess the vascular access. There were several signs reported to be associated vascular stenosis, such as prolonged bleeding, arm edema, and changes in pulse characteristics of the vascular access, but they need a high level of skill and experience from the operator, such as doctors and nurses [2].

Unfortunately, angiographic study is expensive and invasive and needs contrast medium injection. The diagnostic accuracy of color duplex ultrasound in AVF stenosis remains satisfactory, but it is not always available when decreased AVF flow or an acute thrombotic event occurs. Continuous AVF/ AV graft flow monitoring by patient himself is a better option in terms of long-term care. Previous study indicated that stethoscope waveform analysis could be used as an alternative to diagnosis of vascular stenosis. Previous acoustic studies indicated that increased sound intensity, extra sounds, and higher spectral energy infrequency domain, ranging between 300 Hz and 1000 Hz. These findings are associated with the change from laminar flow to turbulent flow downstream the site of vascular stenosis [3]. In addition, phonoangiography studies pointed out that the systolic peak of stenotic segment is higher and narrower in the time domain compared to non-stenotic segment Methods: This is an observational cohort study. We plan to recruit 150 eligible patients with ESRD under regular hemodialysis for at least one year in national Taiwan University Hospital.

1. Data gathering A. The first group of patients (study group) will be recruited from hose referred to endovascular intervention, e.g. angioplasty, due to vascular access dysfunction (n=30). An electronic stethoscope (3M-Littmann) will be used to record digital sound signals at 5 points with 30 seconds each, including AV junction, proximal and distal to both needle puncture sites, respectively. All data will be transferred to a computer for further anlaysis.

B. The second group (validation cohort) will be recruited from patients with regular hemodialysis (HD) and received angioplasty at National Taiwan University Hospital (n=120). Those patients who received angioplasty for AVF or AV graft stenosis will also received follow-up angiography 12 weeks after angioplasty. We will record digital AVF/AV graft sound signals as mentioned above. In addition, blood pressure and AVF/AVG venous pressure, heart rate will be recorded. Diameter stenosis of restenotic segment will also be recorded and a diameter stenosis of 50% is regarded as significant. 2. Post-processing The proposed method is based on the analysis of the envelope of the sound signals. Before, calculating the envelope, all signals will be down-sampled to 2k Hz, followed by a linear-phase, finite impulse response low-pass filter. PCA will then be applied to analysis as described previously [1-4]. PCA (principal components analysis) 3. Analysis Sensitivity, specificity and diagnostic accuracy will be calculated based on each diagnostic criteria or basal function. Area under ROC curve will be calculated based on each diagnostic criteria derived from PCA method.