Кластеризация шумов как способ оценки функции постоянного сосудистого доступа у больных на гемодиализе
Abstract. The study aims to develop an algorithm for assessing spectrographic features of arteriovenous
fistula dysfunction for hemodialysis. Materials and methods. Forty-four patients with native radiocephalic fistula
formed in the distal third of the forearm participated in the research. Using electronic stethoscope, the noise of
arteriovenous fistula was recorded in all patients. 653 spectrograms were analyzed with the method of evaluating
entropy and complexity value. The algorithm of Wishart clustering was applied to detect consistency. Results.
The algorithm developed for audiograms analysis and based on evaluation of “chaotic” and “complex” sound makes
it possible to divide arteriovenous fistula noise into three clusters. Each cluster complies with its possibility of
dysfunction. The algorithm may be applied in developing individual mobile devices to monitor constantly the
condition of arteriovenous fistula.