AbstractIn the medical field, the analysis of the blood sample of the patient is a critical task. Abnormalities in blood cells are accountable for various health issues. Red blood cells (RBCs) are one of the major components of blood. Classifying the RBC can allow us to diagnose different diseases. The traditional time consuming technique of visualizing RBC manually under the microscope is a tedious task and may lead to wrong interpretation because of the human error. The various health conditions can change the shape, texture, and size of normal RBCs. The proposed method has involved the use of image processing to classify the RBCs with the help of Convolution Neural Networks (CNN). The algorithm can extract the feature of each segmented cell image and classify it in various types as Microcytes, Elliptocytes, Stomatocytes, Macrocytes, Teardrop RBCs, Codocytes, Spherocytes, Sickel cell RBCs and Howell jolly RBCs. Classification is done with respect to the size, shape, and appearance of RBCs. The experiment was conducted on the blood slide collected from the hospital and RBC images were extracted from those blood slide images. The obtained results compared with reports obtained by the pathology lab and realized 98.5% accuracy. The developed system provides accurate and fast results due to which it may save the life of patients.
Hierarchy-NMS: Merging Candidate Bounding Boxes for Cerebrospinal Fluid Cell Image Segmentation