Paediatric Sickle Cell Detection using Deep Learning - A Review

摘要

Sickle Red blood cells (SRBC) are defective RBC cells due to aberration from regular disc shaped cells to irregular sickle shaped cells inherited from parents. The characteristics of majority RBC cells changes drastically in its nature by becoming gummy, rigid and crescent or C -shaped cells which makes it tough to traverse through very small blood capillaries and restricting the normal flow of blood by blocking or closing the blood vessels. The major function of RBC cell is to carry oxygen to other parts of the body without which a human cannot perform his routine activities. A red blood corpuscle (RBC) becomes Sickle RBC (SRBC), when the quantity of Haemoglobin A (HgbA), usually present in normal RBC cell, becomes less in quantity when compared to another substance called Haemoglobin S (HgbS), which is present abundantly in abnormal RBC cells, makes normal RBC cells to narrow down and sooner gets destroyed. Most of the normal RBC cells changes its structure and making a person anemic as a result of which it shows symptoms like shortness of breath, fatigue, severe pain throughout the body. Sickle cells are more prevalent among children of Africa, where majority of them born with flawed RBC patterns and even die due to this abnormality. This paper outlines about the various state-of-art deep learning models for early detection of sickle cells among children by classifying defective cells from normal cells and finding faulty image patterns of RBC cells in children by using most advanced neural network technique of Artificial Intelligence called Deep Learning. The digital image of a blood sample after pre-processing is given as an input to deep neural network to automatically identify the large population of sickle cell deficient children belonging to a particular geographic location where majority of children with sickle cell disease can be spotted out easily. According to American Society of Haematology (AHS), Sickle cells can be treated with gene editing therapy called CRISPR.