An intelligent white blood cell detection and multi-classClassification using fine optimal DCRNet

Abstract

ABSTRACT [001] Our Invention “An intelligent white blood cell detection and multi‑class Classification using fine optimal DCRNet “ is a major goal of this research is to develop a Deep Learning (DL) based automatic identification and classification of white blood cells (WBCs) with high accuracy and efficiency. The first phase of research is pre-processing and is accomplished by the Improved Median Wiener Filter (IMWF), which effectively eliminates the noises. The image is resized into a standard image size before filtering. The segmentation process takes place using Color Balancing Binary Threshold (CBBT) algorithm to divide the WBCs and another non-relevant background to improve the classification performance. [002] The features like shape, texture and color of the WBCs are extracted from the segmented images. Finally, the classification takes place, and this is processed by an fine optimal deep convolution residual network (Fine Optimal DCRNet). In addition, the bionic model is introduced to improve classification accuracy. The dataset used in this research is BCCD and LISC datasets. [003] The performance of the proposed model is validated using existing methods of Support Vector Machine (SVM), K-Nearest Neighbor (KNN), VGG-16, VGG-19, ResNet-50, DensetNet-121, DensetNet-169, Inception-V3, InceptionResNet-V2, Xception, MobileNet-224, Mobile NasNet, Tree, Naive Bayes, Ensemble active contour model, k-means clustering and handcraft and deep learned features-scale-invariant feature transform (HCDL-SIFT) in terms of Accuracy, Precision, Recall, Specificity, F-score, Relative Distance Error (RDE), Over-Segmentation Rate (OSR), Under-Segmentation Rate (USR) and Overall Error Rate (OER). For the LISC dataset, the detection model attains an outcome of 99%, 98%, 98%, 99%, 98%, 1.143, 0.0125, 0.056 and 0.125, respectively. For the BCCD dataset, apart from RDE, OSR, USR and OER metrics, the performance is evaluated as 98%, 96%, 98%, 99% and 97%.

Specification

Description:FIELD OF THE INVENTION

[004] Our Invention is related to an intelligent white blood cell detection and multi‑classClassification using fine optimal DCRNet.

BACKGROUND OF THE INVENTION
[005] Research problem Different types of leukocytes play a crucial role in the body's immune response. Abnormalities or imbalances in the number and type of white blood cells can indicate various diseases and medical conditions.

[006] theclassification of leukocytes helps diagnose and monitor these diseases. An increase in the number of specific types of leukocytes, such as Neutrophils, is often a sign of bacterial infections, while an increase in eosinophils can indicate parasitic or allergic reactions.

[007] Accurate WBC classification helps identify the cause of infections. Therefore, in the proposed method, WBC classification is used to detect and diagnose diseases to avoid life-threatening conditions.

[008] Research gap Artifcial Intelligence (AI) models have recently offered infinite applications in the Health Care (HC) industry. The human blood cells are categorized as red blood cells (RBCs) named erythrocytes, WBCs named leukocytes and platelets named thrombocytes.

[009] However, each cell type plays a different role in attacking the disease or infection. However, the classification of WBCs is very challenging because each WBC differs in the nucleus's color, texture, and shape.

PRIOR ART SEARCH
US20190205606A1:Methods and systems for artificial intelligence based medical image segmentation are disclosed. In a method for autonomous artificial intelligence based medical image segmentation, a medical image of a patient is received. A current segmentation context is automatically determined based on the medical image and at least one segmentation algorithm is automatically selected from a plurality of segmentation algorithms based on the current segmentation context. A target anatomical structure is segmented in the medical image using the selected at least one segmentation algorithm.
US20220012890A1: An automated method for segmentation includes steps of receiving at a computing device an input image representing at least one surface and performing by the computing device image segmentation on the input image based on a graph surface segmentation model with deep learning. The deep learning may be used to parameterize the graph surface segmentation model.
WO2018015414A1:Methods and systems for artificial intelligence based medical image segmentation are disclosed. In a method for autonomous artificial intelligence based medical image segmentation, a medical image of a patient is received. A current segmentation context is automatically determined based on the medical image and at least one segmentation algorithm is automatically selected from a plurality of segmentation algorithms based on the current segmentation context. A target anatomical structure is segmented in the medical image using the selected at least one segmentation algorithm.

SUMMARY OF THE INVENTION
[010] The above challenges motivated the development of WBC image recognition with an effective DL model to simultaneously classify WBCs and invisible WBCs such as overlap or partial image exclusion. In the proposed model, segmentation of WBC from images of blood smear is employed using the Color Balancing Binary Threshold Algorithm (CBBT).

[012] This algorithm is applied to the input RGB image, and a softmap is created by combining CMYK and HLS color spaces. The feature extraction stage is essential to enhance the classification's accuracy.

[013] Then, the multi-feature extraction such as shape-circularity, solidity and convexity, color-mean-standard deviation of the nucleus, ROC region and convex hull region from HSV, RGB, LAB and YCrCb color spaces and gray-level co-occurrence matrix (GLCM) texture features are extracted.

[014] The Guided Bistage Feature Selection (GBiFS) model performs the appropriate feature selection process for dimensionality reduction with improved accuracy. Then, the WBC classification is performed using the proposed optimized DL model.

BRIEF DESCRIPTION OF THE DIAGRAM

Fig.1, 2: An intelligent white blood cell detection and multi‑class Classification using fine optimal DCRNet

DESCRIPTION OF THE INVENTION
[015] The classification of WBCs plays a marvelous role in medical image processing. This work uses a Fine Optimal DCRNet to detect WBCs and classify them into multiple classes intelligently.

[016] This approach is carried out in six stages: pre-processing, segmentation, feature extraction, selection, and classification and performance assessment. The pre-processing is done through an Improved Median Wiener Filter (IMWF) for image noise removal. The filtering process occurs after the image is resized to a standard image size that fts the DL classifier.

[017] The CBBT algorithm is used to segment the original image that contains WBCs and other irrelevant background elements, which negatively affect the classification performance.

[018] The multiple feature extraction is carried out based on shape, texture and color. Shape features are obtained through properties like circularity, solidity and convexity, the GLCM provides the texture features and considered color features are ROC region, convex hull region and mean-standard deviation of the nucleus from HSV, RGB, LAB, and YCrCb color spaces.

[019] The essential feature is the selected GBiFS method. Finally, the WBC classification is achieved through Fine Optimal DCRNet, and the loss in Fine Optimal DCRNet is optimized using the FSA bionic model. The performance assessment is carried out based on various performance metrics to evaluate the effectiveness of the approach compared with existing approaches.

[020] BCCD dataset is a small dataset used to detect blood cells in the proposed experimental analysis system. It includes 12,500 enhanced images of blood cells with accompanying cell type labels. The cell types, such as Eosinophils, lymphocytes, monocytes and neutrophils, are grouped in 4 different folders, each folder containing 3000 images.

[021] This dataset is accompanied by another additional dataset containing 410 original images and two additional subtype labels and bounding boxes for each cell in each of these 410 images. Figure 5 shows the example images of the BCCD data set.

[022] The LISC dataset includes 400 sample images of microscope slides and samples from the peripheral blood of 8 normal subjects. The microscopic slides were recorded using a digital camera and saved in the BMP format. The images contain a total of 720 × 576 pixels.

[023] A hematologist classifies these collected color images into normal leukocytes such as basophils, eosinophils, lymphocytes, monocytes, and neutrophils. An expert manually segments the areas related to the nucleus and cytoplasm. Figure 6 shows the sample images of the LISC dataset.

, Claims:I/WE CLAIMS

1. Our Invention "An intelligent white blood cell detection and multi‑class Classification using fine optimal DCRNet" is a major goal of this research is to develop a Deep Learning (DL) based automatic identification and classification of white blood cells (WBCs) with high accuracy and efficiency. The first phase of research is pre-processing and is accomplished by the Improved Median Wiener Filter (IMWF), which effectively eliminates the noises. The image is resized into a standard image size before filtering. The segmentation process takes place using Color Balancing Binary Threshold (CBBT) algorithm to divide the WBCs and another non-relevant background to improve the classification performance. The features like shape, texture and color of the WBCs are extracted from the segmented images. Finally, the classification takes place, and this is processed by a fine optimal deep convolution residual network (Fine Optimal DCRNet). In addition, the bionic model is introduced to improve classification accuracy. The dataset used in this research is BCCD and LISC datasets. The performance of the proposed model is validated using existing methods of Support Vector Machine (SVM), K-Nearest Neighbor (KNN), VGG-16, VGG-19, ResNet-50, DensetNet-121, DensetNet-169, Inception-V3, InceptionResNet-V2, Xception, MobileNet-224, Mobile NasNet, Tree, Naive Bayes, Ensemble active contour model, k-means clustering and handcraft and deep learned features-scale-invariant feature transform (HCDL-SIFT) in terms of Accuracy, Precision, Recall, Specificity, F-score, Relative Distance Error (RDE), Over-Segmentation Rate (OSR), Under-Segmentation Rate (USR) and Overall Error Rate (OER). For the LISC dataset, the detection model attains an outcome of 99%, 98%, 98%, 99%, 98%, 1.143, 0.0125, 0.056 and 0.125, respectively. For the BCCD dataset, apart from RDE, OSR, USR and OER metrics, the performance is evaluated as 98%, 96%, 98%, 99% and 97%.
2. According to claim1# the invention is to a "An intelligent white blood cell detection and multi‑class Classification using fine optimal DCRNet" is a major goal of this research is to develop a Deep Learning (DL) based automatic identification and classification of white blood cells (WBCs) with high accuracy and efficiency.
3. According to claim1,2# the invention is to a first phase of research is pre-processing and is accomplished by the Improved Median Wiener Filter (IMWF), which effectively eliminates the noises. The image is resized into a standard image size before filtering.
4. According to claim1,2,3# the invention is to a segmentation process takes place using Color Balancing Binary Threshold (CBBT) algorithm to divide the WBCs and another non-relevant background to improve the classification performance. The features like shape, texture and color of the WBCs are extracted from the segmented images.
5. According to claim1,2,3# the invention is to aclassification takes place, and this is processed by an fine optimal deep convolution residual network (Fine Optimal DCRNet). In addition, the bionic model is introduced to improve classification accuracy. The dataset used in this research is BCCD and LISC datasets.
6. According to claim1,2,3,4# the invention is to a performance of the proposed model is validated using existing methods of Support Vector Machine (SVM), K-Nearest Neighbor (KNN), VGG-16, VGG-19, ResNet-50, DensetNet-121, DensetNet-169, Inception-V3, InceptionResNet-V2, Xception, MobileNet-224, Mobile NasNet, Tree, Naive Bayes, Ensemble active contour model, k-means clustering and handcraft and deep learned features-scale-invariant feature transform (HCDL-SIFT) in terms of Accuracy, Precision, Recall, Specificity, F-score, Relative Distance Error (RDE), Over-Segmentation Rate (OSR), Under-Segmentation Rate (USR) and Overall Error Rate (OER).
7. According to claim1,2,3,4# the invention is to a the LISC dataset, the detection model attains an outcome of 99%, 98%, 98%, 99%, 98%, 1.143, 0.0125, 0.056 and 0.125, respectively. For the BCCD dataset, apart from RDE, OSR, USR and OER metrics, the performance is evaluated as 98%, 96%, 98%, 99% and 97%.

Documents