A Hybrid Technique for Blood Cell Detection

In pathology, the count of WBC is done manually which yields imperfect results and hematology devices may solve the issues of erroneous results but the cost is very high. The motivation of this work is to provide an automated computer aided system (CAS) which analyses the microscopic blood images. In analysis, the image segmentation is a vital step and if any error occurs in this step inaccurate results may be obtained during cell detection. Initially, the microscopic blood image is converted to a binary image. Next, the segmentation process is employed to detect the blood cells from blood-smeared images. The thresholding estimation method is used to identify the proper blood cells from the gradient image. The gradient based region growing method is applied to detect the boundary of cells so as to prevent any kind of contact between the cells. The watershed transformation is applied over the thresholding based gradient image. The thresholding estimation method is very much efficient to detect the connected cells in gradient image. Finally, Circular Hough (CH) transformation is used to identify and count the different blood cells i.e., WBCs, RBCs, platelets etc from the microscopic blood image. The accuracy of this method is found 91% using a database of 50 blood samples obtained from microscope. Statistical analysis of image quality analysis i.e., SSIM (Structural Similarity Index Measure) of outcome images using the proposed method is also found to be higher than the other existing methods.