Multi-feature-based approach for white blood cells segmentation and classification in peripheral blood and bone marrow images

Abstract

In this paper, we propose a complete automated framework for white blood cells differential count in peripheral blood and bone marrow images, in order to reduce the analysis time and increase the accuracy of several blood disorders diagnosis. A new colour transformation is first proposed to highlight the white blood cells regions; then, a marker controlled watershed algorithm is used to segment the region of interest. The nucleus and cytoplasm are subsequently separated. In the identification step, a set of colour, texture and morphological features are extracted from both nucleus and cytoplasm regions. Next, the performances of a random forest classifier on a set of microscopic images are compared and evaluated. The obtained results reveal high recognition accuracies for both segmentation and classification stage.

Virtual Staining, Segmentation, and Classification of Blood Smears for Label-Free Hematology Analysis

“…While several virtual staining techniques based on a variety of label-free imaging techniques have been presented [29][30][31][32], they are mostly geared toward the staining of tissues for histopathology and are not designed to digitally stain and analyze blood smears. Further, our segmentation method is robust and achieves comparable or even better performance than methods based on stained or pseudocolorized images, without the need for fix-ing and staining the sample [36][37][38] or the need for multispectral imaging [24]. We have presented a simple and robust classification and counting procedure that utilizes cellular and nuclear segmentation masks along with the grayscale images to first exclude dead WBCs and then classify healthy WBCs into five subtypes.…”

“…While several methods for segmentation and classification of WBCs have been proposed, most of them rely on feature extraction or training DNNs using stained images [35][36][37][38][39] or fail to provide an accurate five-part white blood cell differential [15,16,27,28,40]. Here, we present a segmentation method that uses only grayscale images (and is independent of the virtual staining branch of the pipeline) and has very high accuracy, with an average dice score of 0.9899 for cellular segmentation and 0.9718 for nuclear segmentation on an unseen test dataset.…”

White blood cells segmentation and classification using a random forest and residual networks implementation

“…21 This method has shown great promise, in comparison with others, in the classification of histological image data in recent papers, including blood smears. 14,[22][23][24] After classification, the resulting binary image was submitted to a morphological opening (an erosion followed by dilation) by a (5, 5) kernel of ellipsoidal format, followed by a connected components analysis (CCA) with statistics, both functions available on OpenCV library. The opening kernel was an ellipsoidal one, with a size of (5,5), and the parameters of connectivity and ltype, for CCA, as being to 8 and CV 32S, respectively.…”

“…Kutlu et al 13 studied the use of Regional Convolutional Neural Networks (R-CNN), based on AlexNet, VGG16, GoogLeNet, and ResNet50 architectures with full learning and transfer learning. Another article, from Benomar et al, 14 performed segmentation and used color and morphological features extracted from the nucleus and cytoplasm for classification by Random Forest. Another paper published by Zheng et al 15 proposed image segmentation.…”

Modified U‐Net for cytological medical image segmentation

“…Benomar et al 17 presented the different steps of a differential WBC s counting system based on a new color transformation, texture, and shape properties leading to faster and more accurate results. Decorrelation stretch is applied to reveal chromatic characteristics of cells and make the pixels of WBC s more distinguishable.…”