Study On Identification And Classification Of Peripheral White Blood Cells

Peripheral white blood cells(PWBC)play a vital role in the human immune system,it's identification and counting also make a difference in the diagnosis of clinical disease.Manual classification and counting is traditionally used to detect PWBC,it is time consuming and laborious,and the counting results vary from person to person which greatly affecting the clinical diagnosis.The application of digital image processing technology to identify and classify white blood cells,greatly reduce the labor costs,and the results are accurate and efficient,which has a very broad application prospects in the white cell recognition and counting.PWBC are divided into five main kinds of cells in medicine,includes eosinophils,monocytes,lymphocytes,neutrophils and basophils.This paper uses digital image processing technology to process PWBC identification and classification.In this study,the digital images were obtained by electron microscopy at the hospital detection center,and the nucleus region was segmented by threshold segmentation method,and the cytoplasm was divided according to the region of the nucleus.And then for the five types of cells color,morphology,texture and other characteristics of the characteristics of search,a total of 34 important features(including optical density,gray covariance matrix,Hu moment,circularity,eccentricity,centerline and some combination characteristics)were obtained.According to these characteristic parameters and doctors' labels of cells,the random forest method is used to train the classifier,and it can be used to detect and classify the white blood cell samples.Through the comparative test,the random forest classifier can effectively identify and classify white blood cell digital images,and the effective recognition rate of eosinophils,monocytes,lymphocytes,neutrophils and basophils reached 96.08 %,95.65%,98.18%,95.57%,88.46%.