Components Statistics And Analysis Based On Texture Analysis In Gynecologic Microscopic Image

With the development of digital image processing technology, medical image processing and analysis technology is playing an increasingly important role in the field of medical diagnosis, such as brain image segmentation, we can use machine learning methods to judge the possibility of senile dementia; When analyzing the blood cell, we can use the cell microscopic image processing technology to detect the degree of cytopathic effect automatically.Leucorrhoea routine check is generic examination method in gynecological clinical of diagnosing vaginitis. Its check projects include vaginal p H, amine test, microorganism examination, vaginal cleanness and the check of clue cells. The aim of microorganism examination is to diagnose whether there exists mold spores or trichomonas vaginalis, the value of vaginal cleanness is determined by the number of vaginal bacteria, mixed bacteria, epithelial cells and white blood cells. Leucorrhoea routine inspection now mainly relies on manual operation in hospital. The inspection process is time-consuming, the accuracy and reliability of the inspection result is influenced by the subjective factors of physicians. Combined with gynecological vaginal microscopic images and video provided by a company in Shandong province, the system has recognized clues cells, epithelial cells, white blood cells, red blood cells, mold spores, lactobacillus, mixed bacteria and trichomonas vaginalis in the process of microorganism examination, vaginal cleanness and the check of clue cells. At present, the detection algorithm has well practical application.This work includes the following aspects:(1) The preprocessing of the gynecologic secretion microscopic image. We introduce the microscopic features of medical images and analyze the characteristics of each component firstly; then we increase the contrast of components and background by image enhancement.(2) Removing the background of the gynecologic secretion microscopic image. Depending on the characteristics of each component of the image, different image segmentation algorithm is put forward that makes the recognition of each component more easily.(3) The recognition and statistics of each component in the gynecologic secretion microscopic image. In the process of background segmentation, the component can be divided into two kinds: one kind is the epithelial cells and the clues cells, another kind is white cells, red cells, mold spores, lactobacillus and mixed bacteria. Texture recognition algorithm based multi-scale laws energy is proposed to distinguish epithelial cells and clues cells effectively. We use the geometric features and gray level characteristics to detect and recognize components including white blood cells, mold spores, lactic acid bacteria and mixed bacteria.(4) The detection of trichomonas vaginalis in the gynecologic secretion microscopic video. According to the moving characteristics of trichomonas vaginalis, we use the method of frame difference to realize moving target detection firstly, and then recognize trichomonas vaginalis according to the statistical characteristics of trichomonas vaginalis.(5) Completing the calculation of cleanliness. We complete the calculation of cleanliness based on the judgment standard provided by the company.The experimental results show that the segmentation and recognition methods of each component in the microscopic image have achieved well results, and trichomonas vaginalis detection has been realized preliminarily. But there is no effective algorithm to solve issues of cell adhesion segmentation and weak edge segmentation. A new detection algorithm should be proposed to detect trichomonas vaginalis when the background is complex. We hope to solve these problems in the latter study for better application in clinical detection.