LeukoSegmenter: A Double Encoder-decoder Based Network for Leukocyte Segmentation From Blood Smear Images

Segmentation of blood cells is a prerequisite step in automated morphological analysis of blood smear images, cell count determination and diagnosis of various diseases such as leukemia. It is extremely challenging due to different sizes, shapes, morphological characteristics and overlapping of blood cells. Due to its complicated nature, it is generally performed as a sequence of steps. However, sequential segmentation results in restricted accuracy due to cascading of errors that creep during each stage. On the contrary, pixel-wise segmentation of blood cells is a single step task and gives promising results. In this paper, we propose LeukoSegmenter, a double encoder-decoder for precise pixel-wise segmentation of leukocytes from blood smear images. It uses pre-trained ResNet18 based encoders and U-Net based decoders. Feature maps obtained from the first network are utilised as attention maps. These are used as input in conjunction with the original 3-channel image to obtain final mask from the second network. This mechanism allows the latter encoder-decoder pair to focus explicitly on leukocytes and ignore other blood cells and debris, thus improving the segmentation accuracy. Experiments on ALL-IDB1 dataset show that the proposed LeukoSegmenter achieves intersection-over-union score of 94.6827% and Dice score of 97.1987% which is superior than that of state-of-the-art methods.

Leukocyte Image Segmentation Based on Adaptive Histogram Thresholding and Contour Detection

Aims: The proposed method falls into the category of medical image processing. Background: Computer-aided automatic analysis systems for the analysis and cytometry of leukocyte (White Blood Cells, WBCs) in human blood smear images are a powerful diagnostic tool for many types of diseases, such as anemia, malaria, syphilis, heavy metal poisoning, and leukemia. Leukocyte segmentation is a basis of its automatic analysis, and the segmentation accuracy will directly influence the reliability of image-based automatic leukocyte analysis. Objective: This paper aims to present a leukocyte segmentation method, which improves segmentation accuracy under rapid and standard staining conditions. Methods: The proposed method first localizes leukocytes by color component combination and Adaptive Histogram Thresholding (AHT), and crops sub-image corresponding to each leukocyte. Then, the proposed method employs AHT to extract the nucleus of leukocyte and utilizes image color features to remove image backgrounds such as red blood cells and dyeing impurities. Finally, Canny edge detection is performed to extract the entire leukocyte. Accordingly, cytoplasm is obtained by subtracting nucleus with leukocyte. Results: Experimental results on two datasets containing 160 leukocyte images show that the proposed method obtains more accurate segmentation results than their counterparts. Conclusion: The proposed method obtains more accurate segmentation results than their counterparts under rapid and standard staining conditions.

Quick Leukocyte Nucleus Segmentation in Leukocyte Counting

The leukocyte nucleus quick segmentation is one of the key techniques in leukocyte real-time online scanning of human blood smear. We propose a quick leukocyte nucleus segmentation method based on the component difference in RGB color space. By analyzing the captured microscopic images of the peripheral blood smears from the autoscanning microscope, it is found that the difference values between B component and G component (B−G values) in the regions of the leukocyte nuclei and the platelets are much bigger than those in the other regions, even in the regions including the stains. So, the B−G values can segment the leukocyte nuclei and the platelets with an appropriate empirical threshold because the platelets are much smaller than the leukocyte nuclei, so the leukocyte nuclei can be segmented by size filtering. Also, only an 8 bit subtraction operation is performed for the B−G values, and it can improve the leukocyte nucleus segmentation speed significantly. Experimental results show that the proposed method performs well for the five types of leukocyte segmentation with a quick speed. It is very suitable for the real-time peripheral blood smear autoscanning test application. In addition, the five types of leukocytes can be counted accurately.