Computer-Aided Diagnosis Of Eosinophilic Gastroenteritis Pathological Sections

Eosinophilic Gastroenteritis(EG)is a gastrointestinal disease characterized by increased peripheral blood Eosinophil(EOS).The main diagnosis is based on whether the number of Eosinophils exceeds the standard in the pathological section of the digestive tract mucosa specimen.The traditional method of diagnosing EG is to observe the pathological section of the digestive tract biopsy specimen under a light microscope and count the number of EOS cells.Manual counting by pathologists is a highly repetitive task and highly dependented on the experience of the doctor.This study uses an improved watershed algorithm to automatically identify and count EOS cells in pathological sections of the digestive tract mucosa,and assist pathologists in diagnosis.The aim is to reduce the workload of doctors and improve the efficiency of medical diagnosis.The main work arrangements are as follows:(1)In this study,Image acquisition of EG pathological sections was completed at Baoding Children's Hospital.The edge detection algorithm is used to reduce the influence of noise and impurities on the image.The gradient amplitude filtering algorithm is used to enhance the contrast between EOS cells and background cell tissue in the image.Convert RGB color images into YUV space models,to reduce the memory pressure of the computer,and improve the speed of the algorithm.(2)In this paper,thresholds are set in three dimensions of YUV,and threshold segmentation is performed on EG pathological slice images.According to the color characteristics of EOS cells to separated from background cells such as lymphocytes and histiocytes.Based on the area and shape of EOS cells in the images,Mathematical morphology algorithm was used to further isolate the background cell tissue.(3)In this paper,a watershed algorithm is used to segment EOS cells from the biopsy specimens of the digestive tract.Due to the existence of noise or other interference factors,the traditional watershed algorithm is easier to identify pseudo-minimum points,which caused over-segmentation.Based on the traditional watershed algorithm,Throughthe improvement of gradient amplitude filtering distance transformation,background and background labeling and region merging,the over-segmentation phenomenon in the traditional watershed algorithm is improved,and the accuracy of cell recognition and the stability of the algorithm are improved.This paper uses an improved watershed algorithm to identify and count EOS in EG pathological images,and compared it with the pathologist's gold standard.The average accuracy rate is 95.0%.Compared with the traditional algorithm,the relative standard deviation of the accuracy of the improved algorithm is increased from 5.8% to 2.2%,the over-segmentation rate is reduced from13.4% to 3.7%,and the running time of the algorithm is reduced from 40 s to 27 s.(4)In order to facilitate the operation of pathologists,this article designed a GUI interface,including four parts: login interface design,preprocessing interface design,feature extraction interface design,and segmentation counting result interface design.