Applying DCT Combined Cepstrum for the Assessment of the Arteriovenous Fistula Condition

This paper focuses on a comparison of effectiveness of the artificial intelligence techniques in diagnosis of arteriovenous fistula condition. The use of discrete cosine transform (DTC) combined cepstrum in the feature extraction process made it possible to increase the value of classification quality indicators by about 10% (when compared to the previous approach based on averaged energy values in the third octave bands). This paper presents a methodology of extracting features from the acoustic signal emitted by the fistula. The supervised machine learning technique of k-NN, Multilayer Perceptron, RBF Network and Decision Tree C4.5 classifiers was applied to develop the classification model. For this, signals obtained from 38 patients on chronic hemodialysis were used. The results show that the cepstral analysis and obtained features provide an accuracy of above 90% in proper detection of vascular access stenosis.