CNN-Based Approach for Overlapping Erythrocyte Counting and Cell Type Classification in Peripheral Blood Images

Classification and counting of cells in the blood is crucial for diagnosing treating diseases clinic. A peripheral smear method a fast, reliable, robust diagnostic tool examining samples. However, cell overlap during process may cause incorrectly predicted results classifying types. The overlapping problem can occur automated systems manual inspections by experts. Convolutional neural networks (CNN) provide reliable segmentation classification many problems medical field. creating ground truth labels data time-consuming error-prone. This study proposes new CNN-based strategy to eliminate overlap-induced samples accurately determine type. In proposed method, images were divided into sub-images, block block, using adaptive image processing techniques identify CNN was used classify types numbers sub-images. successfully counts erythrocytes determines type with an accuracy rate 99.73\%. have shown that it be efficiently various fields.