Research On The Techniques Of Dual-Threshold Based Peripheral Blood Leukocyte Microscope Image Segmentation

Blood cell morphology analysis is of great significance in the early diagnosis of blood related diseases such as anemia,leukemia,and AIDS.Traditional morphological examinations are mostly carried out by specially trained clinical laboratory personnel classifying and counting the cells in blood smears made from peripheral blood or bone marrow under a microscope.It is tedious and time-consuming.What is worse,operators feel tired easily in that case and it is difficult to get a unified diagnostic result since people have varying levels of specialty experience.Deterioration of today's environment directly or indirectly leads to the increase of the number of people with pathology diseases.As to them,early diagnosis and timely treatment are very important,so the demand for morphological examination becomes more urgent.With the development of modern medical image processing and artificial intelligence technology,fully automatic blood cell morphology analyzer has got more and more attention with its convenient,accurate,and efficient ways of working and more adapting to the trend of the development of remote medical system.The fully automatic blood cell morphology analyzer consists of two parts:hardware and software.The former usually includes a blood smear preparation device,a microscope,a camera,and a computer used for blood cell microscopic image processing.The latter contains algorithms for blood cell image segmentation,feature extraction,blood cell image classification,and counting.Relatively speaking,it is easy to reach a consensus on the selection of hardware devices,so performance of the analyzer mainly depends on the image processing algorithms.Further,since cell segmentation algorithm's accuracy directly affects subsequent feature selection and classification,it plays a vital role in the whole system.This paper focuses on cell image segmentation methods to improve the overall system performance.After a review of commonly used segmentation methods,summarizing their advantages and disadvantages,an adaptive dual-thresholdsegmentation method based on gray image's good foreground and background differentiation and HSV color space's H channel image excellent white blood cell outline retainability abilities is proposed.Two independent thresholds are used for separation of background and cell,red blood cell and white blood cell(WBC)operated respectively in RGB and HSV color spaces.Then,intersect the above two separate threshold segmentation results.By experimenting on the images of a public and free available Acute Lymphoblastic Leukemia(ALL)microscopic image dataset and comparing with two traditional single threshold segmentation methods,it shows that the proposed method can segment the lymphocytes from the images stably and accurately.Accuracy reaches 98%,better than the single threshold segmentation algorithms.What is more,it has a certain robustness even to images with different color characteristics,laying a good foundation for subsequent white blood cell feature extraction,classification,and counting.