Analysis of the vascular sounds of the arteriovenous fistula
Abstract: This licentiate thesis is in the field of biomedical signal processing with main focus on processing of the vascular sounds of the hemodialysis vascular access in patient undergoing hemodialysis. The first paper proposes a feature extraction method based on principal component analysis (PCA) for acoustic detection of venous stenosis. The method exploits the changes on the morphology due to the presence of a stenotic lesion. The results showed high classification accuracy 83%-97% for the different cases studied. The vascular sounds at the anastomosis and downstream of the anastomosis are studied and characterized in the second paper. This is carried out by extracting parameters from the power spectral density and wavelet decomposition. The results showed that the vascular sounds at the anastomosis can be characterized as extra energy in the higher frequency bands (200-1000 Hz). Besides, the energy pattern found with wavelet is very similar to that found in studies of arterial and venous stenosis. In the third paper, the vascular sounds at the anastomosis are proposed as a reference sound to assess the presence or not of stenotic lesions. The squared Euclidean distance and azimuth angle are used as similarity measures between the reference sound and the test sounds. The method was able to correctly discriminate among the three different types of test sounds used. Likewise, the results of the azimuth angle indicate that it is feasible to divide, into regions, the vector subspace spanned by the retained PCA basis functions.
This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.