Corpuscular volume estimation using a small amount of blood via microscope images

Abstract

Blood is an analysis serves as a valuable diagnostic tool that can easily and simply diagnose health, and in particular, red blood cells (RBC) or hematocrit (Hct) and mean corpuscular volume (MCV) can diagnose various diseases. Electrochemical methods have been proposed to find out RBC, Hct, and MCV, but it is difficult to use them for infants or the elderly because they require a large amount of blood. As an alternative, method like blood smear has been suggested to make a diagnosis with a small amount of blood, but it takes a long time to dry the sample. In this study, we present a methodology to make the simple and quick blood analysis. This method uses a microscope image of a small amount of blood, without drying the blood sample. From the image, in order to detect all the blood cells, we adopt a loop algorithm consist of concave point detection, ellipse fitting, and blob detection. To assess the validity and accuracy of the proposed method, an investigation is conducted to examine its correlation with the Sysmex XN-1000, a widely utilized medical reference standard; we show that the correlation coefficients are higher than 0.88, showing more correlation than the method in previous studies by comparing the estimation results. By developing this method, it will be possible to establish a simple quick blood analysis method using microscope images, decreasing a time consumed for the health checkup, saving more lives.