چكيده به لاتين
In this study, some methods for the automatic analysis of images and the determination of the geometric properties of the kidney components are presented. The kidneys consist of about one million filter units called nephrons and each nephron consist of a glomerulus and a tubule. Glomeruli are small filters, and tubules are structurally like a thin tube attached to glomeruli. One of the important indicators in the diagnosis of pathological conditions from the normal cases, as well as the detection of the effect of drugs and injectable substances on the kidney components, is to determine the mean area and thickness of glomerulus. These image features are called morphological features. Currently, in many treatment centers, manual methods are used to quantification of kidney images. These methods, besides being time-consuming, also have significant measurement errors. A total of 94 microscopic images were taken containing single glomerulus from Wistar rats. The proposed algorithm quantify 90 images correctly. Quantification of mean thickness of the glomerular basement membrane and the mean diameter of glomerulus and Bowman's capsule is the general purpose of this thesis. For this, the additional parts on the target object are first identified and eliminated. The base method to calculate the mean thickness is such a way that cross-sectional lines are drawn in different parts of the object, the local thicknesses are calculated and eventually averaged. To calculate the mean diameter, geometric center of glomerulus and Bowman's capsule approximated, the radial lines are plotted from that center, local radiuses calculated and finally averaged. Other methods for thickness and diameter are also presented. Finally, these morphological factors were calculated experimentally by the MOTIC software and the results of the intelligent algorithm were compared whit them. The results of the thickness calculation by base method, have 2.09% error and for glomerulus and Bowman's capsule diameter also have 1.15% and 1.84% errors, respectively, compared to the experimental results. The importance of this comparison is related to the time of the calculations, which reduced hours to several seconds. The results from using the proposed algorithm on images with different shapes demonstrate their satisfactory performance under different conditions. As well as a certificate of high accuracy and speed. As a result, it can be used for accurate and fast morphological measurements for therapeutic applications.
