| The study of the blood cell is important in the field of medical diagnostics. The study can be used to detect and diagnose certain diseases. Normal blood cells should be about the same size and appear as perfectly round circles. But interaction between an antibody and its corresponding antigen results in aggregation of these blood cells (called “clumping”).; At present, human experts manually view images and identify these blood cells and clumped cells. The analysis of clumped cells is a difficult task because clumps do not have any specific shapes and boundaries. The clumped cells are very often blurred and noisy, and cell shapes are often distorted. Thus, this human visual inspection is not only a labor intensive and time-consuming process but it also requires special skill and careful judgment.; An automated blood cell recognition and analysis system was developed in this research. The proposed system automatically identifies the number of normal blood cells and clumped cells from the blood image using the Fast Connected Hough Transformation. Then, a discrimination algorithm developed in this research successfully separates normal blood cells from clumped cells. Finally, the system performs analysis on the clumped cells. The goal of clumping analysis is to calculate the number of cells inside the clumping. To achieve this goal, the Backpropagation Neural Network (BNN) is trained with physical features extracted from the clumping. The testing results showed that the BNN generates a number very close to the actual number of cells for each clumping. Using the BNN built in this research, the analysis of the clumping is automated successfully.; Experimental results with blood cell images show that the proposed system performed the automated blood cell recognition and analysis tasks successfully. The success of this research brings great benefits to medical image processing and thus enhances research of medical diagnostics. |
