“Deep Learning based diagnosis of sickle cell anemia in human RBC”

Abstract

Sickle cell disease is a type of anemia distinguish by irregular erythrocytes that cause blood stream blocking, it is a severe hematological condition that causes people to be treated regularly during their lives and may also result in death. Standard RBC have a spherical form and are compact and resilient, allowing them to travel across narrow capillaries with ease. Irregular RBC’s, on the other hand, have a sickle appearance and are rigid and blunt, allowing them to get trapped in thin blood vessels. Patients will experience discomfort as a result of this, and low oxygen and exhaustion will result. In this research a Deep CNN model, to classy sickle cell disease and data augmentation technique such as flipping, zooming, height and width shift done to get much better accuracy, in this research Idb1 erythrocytes microscopic photographs of blood smears obtained from patients infected with sickle cell, and the dataset is divided into test and train for each classis which are circular, elongated and others. For the classification task five pre trained model are used which are VGG16, VVG19, ResNet50, ResNet101 and Inception V3. Proposed models’ efficiency is shown by the results of the work, which offers better accuracy of the classification.