Различение хаотических и регулярных временных рядов для идентификации состояния артериовенозной фистулы

Вестник кибернетики. 2022. Т. 45. № 1. С. 72–82.

Gromov V., Мазайшвили К. В., Заикин П. В., Николаев Е. Н., Beschastnov Y., Зворыкина Е. И., Parinov A., Neznanov A.

The prevalence of chronic kidney disease is growing every year and is already comparable to such socially significant diseases as hypertension and diabetes mellitus, as well as obesity and metabolic syndrome [1,2]. The standard solution for hemodialysis patients is to create a permanent vascular access in the form of an arteriovenous fistula. However, its use is limited to frequent thrombosis, which entails risks to the patient's life. This paper proposes a method for the automatic classification of the rows of responding, normally and pathologically functioning fistulas. The method is based on the hypothesis that a normally functioning fistula blood flow is laminar, and a pathologically functioning one is turbulent. From a mathematical point of view, we reviewed the problem of distinguishing between regular and chaotic time series. To solve this problem, two methods were considered for distinguishing chaotic series from regular ones. The first involves finding the position of the series on the entropy-complexity plane. Further, the specified position is compared with the identified clusters of values ​​of the set of time series. The second (proposed by the authors) method involves the construction of an object-attribute graph within the framework of the concept of analysis of formal concepts. Both methods used proved to be effective in determining the state of the fistula, the second method was more optimal for detecting thrombosed fistulas.

Keywords: entropy, энтропия, кластеризация, complexity, сложность, clustering, Chaotic time series prediction, хаотические временные ряды, медицинская диагностика, medical diagnostics

Publication based on the results of: Интеллектуальный анализ данных в медицинских информационных системах (2022)