Adaptive Network-Based Fuzzy Inference System for Arteriovenous Shunt Stenosis Screening in Long-Term Hemodialysis Treatment of Patients

Abstract

This study proposed an adaptive network-based Fuzzy inference system (ANFIS) for evaluating arteriovenous shunt (AVS) stenos is in long-term hemodialysis treatment of patients. Due to the frequency spectral varies with the normal blood flow and turbulent flow. The power spectra appear changes in frequency and amplitude with the degrees of AVS stenos is. The proposed diagnosis system consists of signal preprocessing and stenos is degree identification. The Burg autoregressive (AR) method was used to estimate the frequency spectra of phonoangiographic signal and to find the peaky spectra in the region of 0Hz and 800Hz. The frequency spectra showed changes in characteristic frequencies with the degrees of AVS stenos is. The main characteristic frequencies distribute into different bands, overlap bands, or crossing bands. Ambiguous and uncertain information is not easy to identify by human-made decisions. Therefore, ANFIS is designed as an early decision-making model to evaluate the degrees of AVS stenos is. The degrees of stenos is (DOS) were divided into three classes by professional physicians. For 42 long-term follow-up patients, the experimental results show the proposed diagnosis system has greater efficiency for evaluating AVS stenosis.