Selection of optimal parameters of the neuro-fuzzy ANFIS system for classification of Ziehl – Nielsen stained sputum smear images

Author(s):  I.G. Shelomentseva, no, no, Krasnoyarsk State Medical University n.a. prof. V.F. Voyno-Yasenetsky, Krasnoyarsk, Russia, inga.shell@yandex.ru

S.V. Chentsov, Dr., Prof., Siberian Federal University, Krasnoyarsk, Russia

A.N. Narkevich, Krasnoyarsk State Medical University n.a. prof. V.F. Voyno-Yasenetsky,, Krasnoyarsk, Russia

Issue:  Volume 46, № 2

Rubric:  System analysis and processing of knowledge

Annotation:  Tuberculosis (TB) is an important public health issue in this world. The Ziehl – Nielsen method of microscopy is the one of the widely used methods for the diagnosis of tuberculosis. In this paper, we present the results of experiment for calculation of optimal parameters of the neuro-fuzzy classification system for ZN (Ziehl – Nielsen) stained images of sputum smear samples obtained using a light microscope. The authors describe the methods for preprocessing and segmentation of the images. We use the color and shape characteristics of the regions of interest for experimental sample. We use the quantity and parameters of the input and output membership functions for the classical neuro-fuzzy system based on the Sugeno model and size of the range of a single cluster for the subtractive clustering. Authors use the values of mean-square error, regression and accuracy as comparison criteria. The results will be used to build the optimal classifier of ZN (Ziehl – Nielsen) stained images of sputum smear samples obtained using a light microscope.

Keywords:  image processing, tuberculosis bacteria, microscopic, method Ziehl-Nielsen, ANFIS, genfis 1, genfis 2, root-mean-square error, regression, accuracy.

Full text (PDF):  Download

Downloads count:  445