Leukocyte classification based on spatial and spectral features of microscopic hyperspectral images

Observing and identifying blood cells is a direct way for early diagnosis of blood diseases. Traditional blood cell recognition methods are usually time-consuming and laborious tasks for medical staff. This paper proposed an efficient leukocyte recognition method based on microscopic hyperspectral imaging technology. In order to achieve better segmentation performance and further improve the representativeness of features, the sequential maximum angle convex cone algorithm and iterative self-organizing data analysis technique algorithm are combined to segment the leukocytes from microscopic hyperspectral images. In addition, the uniform and rotation invariant local binary pattern is adopted as a textural measurement of the leukocytes. Combined the texture features with shape and spectral features, support vector machine is used to classify the leukocytes into different types. Experimental results show that the proposed method provides higher segmentation and recognition accuracy compared with the existing method. Moreover, the addition of spectral features improves the recognition performance shows the potential diagnosis capacity of microscopic hyperspectral imaging technology.