Blood Cell Detection and Counting via Deep Learning

Abstract

A vital component of clinical medical diagnosis is blood cell count. CNN has devised an effective way of automatically counting blood cells using deep learning-based detection method. Inadequate bounding box alignment and overlapping item recognition are challenges for the CNN detection approach. We suggest a brand-new deep-learning technique called CNN to get over these restrictions. Channel, spatial attention mechanism is incorporated into the feature extraction network resulting in CNN. For residual fusion, CNN can assist the network in increasing detection accuracy by replacing the original feature vector and employing the filtered and weighted feature vector. The experimental results show that the typical CNN network may improve blood cell count detection performance without adding too many extra parameters, where the accuracy of identifying cells (RBCs, WBCs, and platelets) has been done.