Acoustic detection of stenosis in hemodialysis vascular access based on the morphology

In this paper, the method proposed in ¡Error! No se encuentra el origen de la referencia. was extended by introducing two new features: F_m and B_m. These features use time scaling to take into account all the beat morphology and to compensate due to the differences in heart rates from patient to patient. The database consists of before and after angioplasty recordings as well as reference recordings. The assessment of the new features was carried out at two levels: intra-patient (by patient), and inter-patients (all patients). Moreover, the performances of the new features were compared to the feature proposed in ¡Error! No se encuentra el origen de la referencia. The results at the intra-patient level showed that the accuracy of SVM classifier for F_m and B_m parameters, on average, increased by 4.2% and 4% respectively. At the inter-patients level the increase was 10% for F_m and 6% for B_m when before angioplasty and reference recordings were compared. In addition, statistical significance was obtained for first principal component analysis coefficient.