Medical Diagnosis by Analysis of Blood Cell Images

This research work aims to investigate the required image processes to extract the features of the red blood cells, and to utilize these features to recognize the normal blood cells from some considered types of anemia infected blood cells. Two sets of features (i.e., geometrical and textural) were suggested, and their discrimination capabilities have been tested. The suggested blood cell recognition scheme consists of three stages. The first stage implies all image processing steps required to determine the background/target colors and to isolate the cell area (target) from the surrounding. Then, the external boundary pixels of the cell cut-out are traced. In the second stage the trace points are used to determine some adopted geometrical features (like; Fourier descriptors, aspect ratio, moments) which have been used to describe the shapes of the blood cells. Also, some textural features are computed to evaluate the spatial color variation within the blood cell. A comprehensive analysis was performed to determine the capability of the considered geometrical and textural features to recognize the types of blood cells.