White Blood Cells Image Classification Using Deep Learning with Canonical Correlation Analysis

doi:10.1016/j.irbm.2020.08.005

IRBM

Volume 42, Issue 5, October 2021, Pages 378-389

https://doi.org/10.1016/j.irbm.2020.08.005 Get rights and content

Highlights

•
Canonical correlation analysis (CCA) employed using CNN-LSTM network architecture.
•
CCA extracts overlapping and multiple nuclei patches from blood cell images.
•
Addition of CCA in combination with CNN-LSTM shows better classification accuracy.

Abstract

White Blood Cells play an important role in observing the health condition of an individual. The opinion related to blood disease involves the identification and characterization of a patient's blood sample. Recent approaches employ Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and merging of CNN and RNN models to enrich the understanding of image content. From beginning to end, training of big data in medical image analysis has encouraged us to discover prominent features from sample images. A single cell patch extraction from blood sample techniques for blood cell classification has resulted in the good performance rate. However, these approaches are unable to address the issues of multiple cells overlap. To address this problem, the Canonical Correlation Analysis (CCA) method is used in this paper. CCA method views the effects of overlapping nuclei where multiple nuclei patches are extracted, learned and trained at a time. Due to overlapping of blood cell images, the classification time is reduced, the dimension of input images gets compressed and the network converges faster with more accurate weight parameters. Experimental results evaluated using publicly available database show that the proposed CNN and RNN merging model with canonical correlation analysis determines higher accuracy compared to other state-of-the-art blood cell classification techniques.

Graphical abstract

Introduction

Blood cells primarily consist of red blood cells (RBCs), white blood cells (WBCs) and platelets [1]. In blood, WBCs contribute significant role in the human body's immune system, moreover, it is also referred to as an immune cell. WBC defends the body against epidemic diseases and extrinsic invaders. Usually, hematologists use coarse knowledge in medical application to classify white blood cells into two groups (a) granular cells: neutrophil, eosinophil, basophil and (b) non-granular cells: monocyte and lymphocyte [2]. Doctors consistently use this primitive data as standards for deciding the seriousness of blood disease. Thus, the research of white blood cell classification has become an important and essential need in the medical analysis.

Each kind of white blood cell is defined with a particular defending role to resist against foreign species. The Neutrophil is one of the main white blood cells which protects against bacteria. Eosinophils destroy parasites and perform a task in allergies whereas the basophils function in allergic reactions. Monocyte enters the cell to destroy the damaged tissues from the body. Lymphocytes are complicated cells to balance cell-mediated immunity. Lymphocytes are totally different from the other WBC, as a result, they memorize invasive micro-organisms and viruses. Fig. 1, Fig. 2.

Different blood cell classification algorithms are primarily focused on white blood cell classification. According to WBC features, appearances, textures, patterns and various attributes extracted by the feature engineering technique in deep learning plays a principal role in image classification. At present, image processing techniques have introduced for classification, segmentation, feature extraction, and identification [4]. The model previously introduced to classify the WBCs images by hematology analyzer was manual process and time-consuming. Also, it introduces errors in detection. Computer-aided diagnostic machine learning algorithms based on image processing like k-means clustering, decision tree, support vector machine (SVM), and merging of CNN and RNN [3] are compared in Table 1. These models can be enhanced further to classify and to produce higher accuracy rates.

The objective of this research is to improve the classification accuracy employing multiple nuclei patch by overlapping features of cells simultaneously, rather than extracting single cell nuclei patch from WBC subtypes. Deep Learning model learns, trains, validates and regenerates the features from multiple images by itself. This innovative approach of overlapping nuclei performed using the canonical correlation analysis techniques helps to extract the features and patches from input images. CCA method helps to enhance the recognition rate, reduce computational time and overfitting problems [5]. Further, CCA trained images are applied to CNN which converges with RNN for better classification accuracy.

This paper is organized as follows: In section 2, background review and some related works in the field is discussed. Section 3 presents the techniques and methodology followed to accomplish the research goals. In section 4, the experimental results are presented along with the discussion. The article is concluded in last section 5.

Section snippets

Review of related literature

In this section, various algorithms of white blood classification are briefly described. Rosyadi et al. introduced an analysis that has the capability of WBC image classification by considering blood smear samples using the digital microscope [12]. The authors further adopted the Ostu threshold technique and KNN grouping method for classification. Depending on their investigation it was finalized that after implementation of k-means clustering to classify WBC, configured feature generated

Methodology

This section illustrates the approach for the classification of WBC image samples utilizing the canonical correlation analysis method. The flowchart for the proposed algorithm is depicted in Fig. 3. In this work, input image feature extraction is performed with the CCA technique to obtain a higher classification rate.

The most important and powerful addition in the network is the CCA, that has achieved better detection accuracy and introduced by Hotelling [21]. In the CCA network images are

Experimental results and discussion

The canonical correlation analysis approach is applied over the entire dataset of blood cell images before undergoing training and validating process through convolutional neural network and recurrent neural network. Instead of single cell images, it considers multiple cell images due to which higher features are extracted and results in high accuracy using CCA technique.

Conclusion

This article presents canonical correlation analysis based deep learning architecture by combining CNN and LSTM for blood cell image classification task. CCA extracts variety of overlapped features from input image thereby enhancing the accuracy rate as compared to other similar deep learning algorithms. Fine-tuning and transfer learning is used to train the WBCs dataset. The results evaluated from proposed model have been satisfactory, showing better precision, sensitivity, specificity and

Human and animal rights

The authors declare that the work described has been carried out in accordance with the Declaration of Helsinki of the World Medical Association revised in 2013 for experiments involving humans as well as in accordance with the EU Directive 2010/63/EU for animal experiments.

Informed consent and patient details

The authors declare that this report does not contain any personal information that could lead to the identification of the patient(s).

Funding

This work did not receive any grant from funding agencies in the public, commercial, or not-for-profit sectors.

Author contributions

All authors attest that they meet the current International Committee of Medical Journal Editors (ICMJE) criteria for Authorship.

CRediT authorship contribution statement

A.M. Patil: Conceptualization, Data curation, Formal analysis, Methodology, Resources, Software, Validation, Writing - original draft. M.D. Patil: Formal analysis, Methodology, Project administration, Resources, Supervision, Writing - review & editing. G.K. Birajdar: Conceptualization, Data curation, Formal analysis, Methodology, Supervision, Validation, Visualization.

Declaration of Competing Interest

The authors declare that they have no known competing financial or personal relationships that could be viewed as influencing the work reported in this paper.

References (31)

Xinghao Yang et al.
Canonical correlation analysis networks for two-view image recognition
Inf Sci
(2017)
Shiqi Yu et al.
Convolutional neural networks for hyperspectral image classification
Neurocomputing
(2017)
Andrea Acevedo et al.
Recognition of peripheral blood cell images using convolutional neural networks
Comput Methods Programs Biomed
(2019)
I.T. Young
The classification of white blood cells
IEEE Trans Biomed Eng
(1972)
S.F. Bikhet et al.
Segmentation and classification of white blood cells
(2000)
G. Liang et al.
Combining convolutional neural network with recursive neural network for blood cell image classification
IEEE Access
(2018)
Alex Krizhevsky et al.
Imagenet classification with deep convolutional neural networks
Adv Neural Inf Process Syst
(2012)
Samidha S. Kulkarni et al.
Classification of RBC and WBC in peripheral blood smear using KNN
Paripex - Indian J Res
(2013)
D.M. Ushizima et al.
Support vector machines applied to white blood cell recognition
K.A.S.A. Daqqa et al.
Prediction and diagnosis of leukemia using classification algorithms

A. Gautam et al.

Automatic classification of leukocytes using morphological features and naive Bayes classifier

Jianwei Zhao et al.

Automatic detection and classification of leukocytes using convolutional neural networks

Med Biol Eng Comput

(2017)

Md. Zahangir Alom et al.

Recurrent residual convolutional neural network based on U-Net (R2U-Net) for medical image segmentation

T. Rosyadi et al.

Classification of leukocyte images using k-means clustering based on geometry features

L. Malihi et al.

Malaria parasite detection in giemsa-stained blood cell images

Cited by (85)

Harnessing deep learning for blood quality assurance through complete blood cell count detection
2024, e-Prime - Advances in Electrical Engineering, Electronics and Energy
Accurate evaluation of blood cells plays a pivotal role in assessing immune system functionality and diagnosing various disorders. While complete blood counts (CBCs) can be obtained rapidly using cell counters, manual blood smear inspection remains crucial for patient monitoring and verification of results. However, this manual analysis is labor-intensive, time-consuming, and prone to errors. To overcome these challenges, we propose an automated solution combining advanced image-processing techniques and deep learning algorithms to automatically recognize complete blood cells in blood smear images. This study introduces an enhanced CNN method for automated blood cell identification and categorization, addressing the challenges associated with manual blood smear inspection in patient monitoring and result verification. The proposed solution combines advanced image-processing techniques and deep learning algorithms to automatically recognize complete blood cells in blood smear images. The approach involves preprocessing input images and harnessing the power of deep neural networks to enhance predictions and corrections. A novel method is also introduced, effectively utilizing the interaction between model outcomes and geographical information to improve classification quality. The proposed model achieves impressive metrics through extensive training and validation, including 90.18 % accuracy, 91 % precision, 90 % recall, and 88 % F-score, surpassing conventional Computer-Aided Diagnosis (CAD) systems in clinical laboratories. The utilization of independent creation and evaluation of neural algorithms with the Adam optimizer enhances the robustness and reliability of the proposed method. The results demonstrate the potential of the approach in clinical and diagnostic labs, offering accurate and efficient automated CBC recognition. The study outlines the plan to integrate the model with a blockchain-based blood supply management system. This integration aims to ensure a secure and efficient delivery of high-quality blood during transfusions, enhancing patient safety and care. By relieving hematologists from labor-intensive and time-consuming tasks, the technology presented in this study paves the way for a future where precision, excellence, and optimal patient outcomes converge seamlessly in the field of blood cell identification and categorization.
Deep integrated fusion of local and global features for cervical cell classification
2024, Computers in Biology and Medicine
Cervical cytology image classification is of great significance to the cervical cancer diagnosis and prognosis. Recently, convolutional neural network (CNN) and visual transformer have been adopted as two branches to learn the features for image classification by simply adding local and global features. However, such the simple addition may not be effective to integrate these features. In this study, we explore the synergy of local and global features for cytology images for classification tasks. Specifically, we design a Deep Integrated Feature Fusion (DIFF) block to synergize local and global features of cytology images from a CNN branch and a transformer branch. Our proposed method is evaluated on three cervical cell image datasets (SIPaKMeD, CRIC, Herlev) and another large blood cell dataset BCCD for several multi-class and binary classification tasks. Experimental results demonstrate the effectiveness of the proposed method in cervical cell classification, which could assist medical specialists to better diagnose cervical cancer.
LMNS-Net: Lightweight Multiscale Novel Semantic-Net deep learning approach used for automatic pancreas image segmentation in CT scan images
2023, Expert Systems with Applications
In the study and research of medical images, the sharp and smooth pancreatic segmentation challenge is a critical and challenging one. The most widely utilized and effective technique for pancreatic segmentation with smooth and precise results is a proposed LMNS-net deep learning, bottom-up approach. The proposed LMNS-net is used to automatically segment the pancreas in clinical abdominal computed tomography (CT) images. For the segmentation of the acute pancreas using several angles of CT scans, such as coronal, axial, and sagittal, a proposed LMNS-net model is used. In the LMNS-Net model, 12 layers are used with 4 convolution layers. LMNS-Net model is used for many organ segmentation from CT scans clinical images with high accuracy. The lightweight multiscale block is used in the proposed approach which is aggregating the required feature only so unused information dropout at the convolution layer. The computation time-period is reduced as related to the state-of-art. Validation is 99.78% and loss values vary from 1 to 0 only. The LMNS-Net model achieved a dice similarity index score up to 88.68±57.49%. The LMNS-Net model takes 1 to 3 s for the segmentation of medical CT images in the testing process. The LMNS-Net model takes very less time for testing purposes as compared to other approaches. Top-down approaches are not useful in the detection of pancreatic cancer images from CT scan images. In Bottom-Up approaches, the LMNS-Net is useful to detect accurate sharpness of the pancreas and kidney, the pancreatic cancer-affected area within less time only. In medical image segmentation, the time of training deep learning approach is a key constraint. “VGG-16” and “VGG-19” are taking too much time, so more time is required for disease diagnosis. The training time of the LMNS-Net model is reduced as compared to the State-of-Arts.
Automatic base-model selection for white blood cell image classification using meta-learning
2023, Computers in Biology and Medicine
Healthcare has benefited from the implementation of deep-learning models to solve medical image classification tasks. For example, White Blood Cell (WBC) image analysis is used to diagnose different pathologies like leukemia. However, medical datasets are mostly imbalanced, inconsistent, and costly to collect. Hence, it is difficult to select an adequate model to overcome the mentioned drawbacks. Therefore, we propose a novel methodology to automatically select models to solve WBC classification tasks. These tasks contain images collected using different staining methods, microscopes, and cameras. The proposed methodology includes meta- and base-level learnings. At the meta-level, we implemented meta-models based on prior-models to acquire meta-knowledge by solving meta-tasks using the shades of gray color constancy method. To determine the best models to solve new WBC tasks we developed an algorithm that uses the meta-knowledge and the Centered Kernel Alignment metric. Next, a learning rate finder method is employed to adapt the selected models. The adapted models (base-models) are used in an ensemble learning approach achieving accuracy and balanced accuracy scores of 98.29 and 97.69 in the Raabin dataset; 100 in the BCCD dataset; 99.57 and 99.51 in the UACH dataset, respectively. The results in all datasets outperform most of the state-of-the-art models, which demonstrates our methodology’s advantage of automatically selecting the best model to solve WBC tasks. The findings also indicate that our methodology can be extended to other medical image classification tasks where is difficult to select an adequate deep-learning model to solve new tasks with imbalanced, limited, and out-of-distribution data.
White blood cell automatic classification using deep learning and optimized quaternion hybrid moments
2023, Biomedical Signal Processing and Control
The structural characteristics of white blood cells (WBCs) can provide important information about human health status. For this reason, over the past four decades, researchers have sought ways to automate the classification and morphological analysis of white blood cells. In the present study, we propose a new method to classify white blood cells into four types: neutrophils, lymphocytes, monocytes, and eosinophils by combining the hybrid discrete moment quaternion approach with deep learning. The proposed classification method is generally divided into two main phases: the first one is called the preprocessing phase, which is devoted to the computation of the image moments to be classified using a new quaternion Meixner-Charlier hybrid moment that is characterized by the following parameters $α, β, a n d φ .$ In addition, the grey wolf optimization algorithm is used to guarantee high classification accuracy by optimizing the local parameters of the new quaternion Meixner-Charlier hybrid moments. The classification phase, which is the second phase of our proposed convolutional neural network model, is dedicated to introducing image moments. We present various graphical measurements, including receiver operating characteristics, precision-recall curves, and a confusion matrix. To demonstrate the effectiveness of our classification method, we also employ numerical measures such as the F1 score, loss, and accuracy. Our results indicate that the cell types of neutrophils were determined with a 98.53% accuracy rate, lymphocytes with a 98.16% accuracy rate, monocytes with a 97.43% accuracy rate, and eosinophils with a 97.43% accuracy rate.
White blood cells classification using multi-fold pre-processing and optimized CNN model
2024, Scientific Reports

View all citing articles on Scopus

View full text

Original ArticleWhite Blood Cells Image Classification Using Deep Learning with Canonical Correlation Analysis

Highlights

Abstract

Graphical abstract

Introduction

Section snippets

Review of related literature

Methodology

Experimental results and discussion

Conclusion

Human and animal rights

Informed consent and patient details

Funding

Author contributions

CRediT authorship contribution statement

Declaration of Competing Interest

Inf Sci

Neurocomputing

Comput Methods Programs Biomed

The classification of white blood cells

IEEE Trans Biomed Eng

Segmentation and classification of white blood cells

Combining convolutional neural network with recursive neural network for blood cell image classification

IEEE Access

Imagenet classification with deep convolutional neural networks

Adv Neural Inf Process Syst

Classification of RBC and WBC in peripheral blood smear using KNN

Paripex - Indian J Res

Support vector machines applied to white blood cell recognition

Prediction and diagnosis of leukemia using classification algorithms