Segmentation and detection of sickle cell red blood image

The most common and dangerous hereditary disease that affect red blood cells (RBC) is sickle cell anaemia due to its morphological characteristics of the cells and caused episodes of pains to the affected individual. This work proposed algorithms in two phase, firstly is to compare segmentation systems such as watershed, edge detection, laplacian of Gaussian and Otsu thresholding on sickle cell anaemia blood smear images and secondly is to detect the presence of cell abnormalities in blood smear images using labelling method by considering eccentricity and form factor features. The RBCs of sickle cell anaemia patient have several abnormalities apart from the sickle shape that will guide medical practitioners on the severity level. The major requirement of the system is to get accurate thresholding level in order to detect the abnormalities of sickle cell anaemia patients for excellent management of the affected individuals to reduce episodes of crises. The phase one proved Otsu thresholding with the highest accuracy, sensitivity and specificity of 93%,94% and 80% respectively by considering 30 blood smear images while the classification gives accuracy, sensitivity and specificity of 88%,93% and 50% respectively