A self-adaptive approach for white blood cell classification towards point-of-care testing

Abstract

As important immune cells in the human body, white blood cells play a very significant role in the auxiliary diagnosis of many major diseases. Clinically, changes in the number and morphology of white blood cells and their subtypes are the prediction index for important, serious diseases, such as anaemia, malaria, infections, and tumours. The application of image recognition technology and cloud computing to assist in medical diagnosis is a hot topic in current research, which we believe have great potential to further improve real-time detection and improve medical diagnosis. This paper proposes a novel automatic classification framework for the recognition of five subtypes of white blood cells, in the hope of contributing to disease prediction. First, we present an adaptive threshold segmentation method to deal with blood smear images with nonuniform colour and uneven illumination. The method is designed based on colour space information and threshold segmentation. After successfully separating the white blood cell from the blood smear image, a large number of features, including geometrical, colour, and texture features are extracted. However, redundant features can affect the classification speed and efficiency, and in view of that, a feature selection algorithm based on classification and regression trees (CART) is designed to successfully remove irrelevant and redundant features from the initial features. The selected prominent features are fed into a particle swarm optimisation support vector machine (PSO-SVM) classifier to recognise the types of white blood cells. Finally, to evaluate the performance of the proposed white blood cell classification methodology, we build a white blood cell data set containing 500 blood smear images for experiments. The proposed methodology achieves 99.76% classification accuracy, which well demonstrates its effectiveness.