Image enhancement and feature extraction techniques for intelligent thalassemia screening procedures

Abstract

Thalassemia is an inherited blood disease that effects the production of red blood cells (RBC) in the body. The deficiency of RBC that constitutes 99% of blood cells will affect their main function as oxygen carrier. The morphological features of RBC play a crucial role in medical diagnosis. Currently, the microscopic investigation to identify the present of any thalassemia cells is performed manually by haematologists through visual identification under a light microscope. However, the manual procedure yields inaccurate results, labour-intensive and time-consuming since it is highly dependent on the haematologists experience and skill. Thus, the main objective of this research is to develop an intelligent thalassemia screening procedure based on blood samples. Essentially an efficient method using the image processing techniques including image enhancement, segmentation and feature extraction have been constructed in order to obtain a fully segmented RBC. There are five contrast enhancement techniques have been applied to the original RGB image, separately and two techniques were selected to be used in the proposed procedure that are Dark Contrast and Partial Contrast techniques. Then, the segmentation of blood cells proceed with the conversion of HSI color space before the image being processed using moving k-mean clustering and median filter techniques. Good segmentation performance for blood cell has been obtained from the combination of dark contrast technique and intensity component. Then, fully segmented RBC image was obtained after the unwanted components such as overlapping RBC, WBC and platelet were successfully removed using seed region growing technique. Next, three main features were extracted from the individual RBC that are simple shape, color and complex shape based features.