An automatic red blood cell counting method based on spectral images

Blood cell analysis, including blood cell counting, is the key point for modern pathological study as well as medical diagnosis. Taking into account both resources and environment of the medical research, analyzing blood cells under the microscope, instead of dedicated blood cell analyzer, provides a more intuitive and convenient way for research uses. This paper aims to provide a method to count red blood cells (RBCs) automatically by analyzing blood cell images collected from a microscopic hyperspectral imaging system. The classification algorithms—spectral angle mappings (SAMs) and support vector machines (SVMs) are used to segment blood cell image. In order to identify RBCs in the image, a standard RBC model has been built to match RBCs in the segmentation results based on SAM classification algorithm. RBC counting results are therefore obtained from the identification and the counting accuracy reaches about 93%. For the sake of higher precision, an improved algorithm, using segmentation results based on SVM classification algorithm to screen the previous matching results, is proposed and the counting accuracy increases to about 98% after applying the improved algorithm.