Patient-Centric Diagnosis of Acute Leukemia: A Machine Learning Approach Utilizing Flow Cytometry Data

doi:10.21203/rs.3.rs-5266356/v1

As cancer incidence rises due to lifestyle factors, accurate and cost-effective leukemia diagnosis is crucial in medical diagnostics. Early identification of malignant leukocytes remains challenging, prompting this research to employ machine learning (ML) techniques. Using a flow cytometry test of blood cells which including Forward Scatter Signal, Forward Scatter Pulse Width Signal, Side Scatter Signal, and Side Fluorescence Signal, various ML algorithms (SVM, K-Nearest Neighbor, AdaBoost Algorithm, Logistic Regression, Decision Tree, Random Forest) are applied for individual patient diagnosis. This study underscores the significance of ML in processing leukemia flow cytometry test signals, enhancing accuracy, reducing diagnosis time, and offering cost-effective and safer diagnostic services. By utilizing ML-based approaches, clinicians and laboratory experts can potentially enhance the efficiency of Leukemia detection and classification. This article provides an in-depth review of current machine learning models used for detection and classification of Leukemia, highlighting their methodologies for discrete data of flow cytometry test, shedding light on their potential benefits and challenges. Overall, this research contributes to the ongoing efforts to improve Leukemia diagnosis through innovative and advanced computational approaches.

Acute Leukemia

Flow Cytometry test

Machine Learning

Leukemia, a deadly cancer, involves the rapid proliferation of abnormal white blood cells, marking it as a malignant hematologic tumour[1]. This condition disrupts the normal balance of red blood cells, white blood cells, and platelets in the body, critical for oxygen transport, infection defence, and bleeding control[2].

White blood cells are vital for the immune system, defending against infections and diseases. Accurate disease diagnosis relies on classifying white blood cells [3] based on cytoplasm composition. Leukemia is categorized as lymphoblastic or myelogenous depending on the affected white blood cell type. Acute Myeloid Leukemia (AML) occurs when granulocytes and monocytes are influenced, whereas Acute Lymphoblastic Leukemia (ALL) is diagnosed when lymphocytes are affected[4]. This categorization is crucial for determining the specific type of Leukemia and guiding appropriate treatment strategies. Haematologists can differentiate various types of Leukemia using microscopic images of correctly stained slides in cell transplant laboratories. While some Leukemia types are more easily identifiable than others through this method, additional equipment is often required for a comprehensive diagnosis. Early diagnosis of Leukemia is challenging due to the mild nature of its initial symptoms, posing difficulties for researchers, doctors, and haematologists. In recent years, machine learning (ML) and computer-aided diagnostic techniques[43] have been increasingly utilized for laboratory image analysis in the field of haematology. These innovative approaches aim to address the challenges of late Leukemia diagnosis and accurately determine its subgroups. By analysing blood smear images, can be diagnosing, differentiating, and quantifying cells in various Leukemia types, enhancing accuracy and speed in medical diagnostics[5, 6]. Flow cytometry is a crucial technique in Leukemia diagnosis and monitoring, leveraging the principles of light scatter and fluorescence to identify abnormal cells. By staining cells with fluorescent antibodies specific to cellular markers and passing them through a laser beam, flow cytometry captures variations in light scatter and fluorescence emissions to analyse cell properties. Forward Scatter (FSC) assesses cell size by measuring light scatter in the direction of the laser, with larger cells producing stronger FSC signals. Side Scatter (SSC) evaluates internal cell complexity, with granular or complex cells reflecting more light sideways, indicating higher SSC signals. Additionally, Forward Scatter Pulse Width (FSC-W) and Side Scatter Fluorescence (SSC Fluorescence) offer deeper insights into cell shape and the detection of specific granules or structures, respectively, enhancing the capability to distinguish between different cell types and states, critical for effective Leukemia assessment.

In recent literature, significant strides have been made in the development of automated systems for detecting Leukemia from microscopic images. Researchers have successfully applied advanced technologies such as image processing, computer vision, and machine learning to create efficient and objective tools. These automated systems analyse cellular morphology, distinguishing normal and abnormal cells associated with Leukemia. The work not only enhances the speed and consistency of diagnosis but also allows for early detection and provides a scalable solution for processing large datasets. The continuous refinement of algorithms and integration into clinical practices holds promise for improving Leukemia diagnostics and ultimately contributing to better patient outcomes.

Kokeb et al.[7] introduced a methodology enhancing instance segmentation accuracy for nuclei detection through pre-processing steps and K-means clustering, aimed at identifying various types of Leukemia cells via an SVM algorithm. They employed metrics such as size, shape, and area for cancer cell segmentation[8]. By extracting features from images and utilizing classifiers built on SVM and GBDT, they demonstrated that a GBDT-based classifier significantly outperforms the SVM model, offering a versatile solution applicable across different platforms. Madhukar et al.[9] developed a decision support tool designed to assist in selecting the appropriate treatment intensity for childhood Acute Lymphoblastic Leukemia (ALL). The tool utilizes K-means clustering for segmenting data based on shape and texture features of the cells. This approach facilitates a more informed and precise decision-making process in determining the most effective treatment plans for young patients with ALL, highlighting the potential of machine learning techniques in enhancing patient care strategies. E.Fathi, et.al. [10] use of an adaptive neuro-fuzzy inference system (ANFIS) is well-suited for distinguishing between patient and non-patient individuals due to its ability to adapt and learn from data uncertainties. For more complex distinctions, such as between acute lymphoblastic Leukemia (ALL) and acute myeloid Leukemia (AML), feature reduction is essential to focus the classification process on key distinguishing attributes, improving both accuracy and model performance. S. S. Al-jaboriy et.al.[11] proposed an automatic leukocyte cell segmentation method utilizing machine learning and image processing techniques was developed, achieving a blast cell segmentation accuracy of 97The system uses 4-time feature analysis and artificial neural networks (ANN) to get the best results on blastema cells even under different lighting conditions. Francis E et.al.[12A computational method has been developed for the diagnosis of leukemia in microscopic blood. Perform histogram-based thresholding of the S component of the HSV color space and then perform morphological erosion for image segmentation. F. E. Al-Tahhan et.al.[13] presented quick and easy classification aims to distinguish all subtypes. Use a segmentation algorithm to capture the main features of the image dataset (10 geometric features). All possible variants of the extracted results are used to train and test some powerful classifiers such as KNN with different functions, SVM and ANN with different kernels. Mohapatra et.al.[14] implemented image segmentation and integrated classification systems are applied to blood microscopic images for early diagnosis of all subtypes. The extracted features are shape, contour, fractal, texture, color and Fourier descriptors. Although these methods give good results with 95% accuracy, they use different parameters, which requires a lot of time and effort. Dumyan, S. et.al.[15] using the developed method for cancer diagnosis using ANN classification, the classification accuracy is estimated to be 97.8%. Use features like patterns, texture properties, statistics and time constants. Karthikeyan, T.et.al.[16] suggested microscopic image segmentation using fuzzy C-means for Leukemia diagnosis. A classification is made by SVM after using Gabor texture extraction method, and the performance accuracy of this technique is of the rate 90%.In the realm of diagnosing and classifying acute lymphatic Leukemia (ALL) with machine learning, most approaches rely on extracting a large number of features to ensure accuracy, which often leads to overfitting and increased noise in the models. This complexity can extend the effort and time needed for the classification process. Nonetheless, a minority of studies have explored the use of algorithms that require fewer features, aiming for streamlined, efficient models that avoid overfitting while maintaining diagnostic precision. These approaches represent an important shift towards developing simpler, faster, and potentially more reliable diagnostic tools for Acute Leukemia[17]. Image segmentation is considered one of the most important steps in computer vision and image processing technology. In this method, the input image is split into multiple objects with the same set of objects to extract the region of interest (ROI) [18] Researchers have used various segmentation methods to detect cancer cells, such as region-based segmentation, growing region segmentation, edge detection segmentation, and cluster-based segmentation. [19][17]. Computer-assisted analysis of blood cells consists of several steps, including preparation, distribution of blood samples from smears, identification of blood cells, and distribution of blood cells. [20, 21]. Accurate detection of the region of interest (ROI), specifically blast cells, is crucial for the efficient classification of leukocyte cells. It's important to note that the majority of classification errors in white blood cells (WBCs) stem from inaccuracies in the segmentation phase. This highlights the critical role of precise segmentation in ensuring the reliability of automated blood cell analysis systems[22, 23, 24].The segmentation of leukocyte cells from microscopic images is challenging due to several factors: overlapping cells, variable image contrast, diversity in cell shapes and sizes, low contrast between objects and the background, and image noise. These difficulties are compounded in the case of white blood cells (WBCs) because of their intricate nuclear structure, making WBC segmentation particularly complex[25]. Researchers have used many different methods to identify tumors using microscopic blood, such as edge segmentation detection, clustering, thresholding, regional classification, and hybrid methods. More importantly, k-means clustering and edge-based segmentation are the most popular methods for cell segmentation in peripheral blood smears [26]. Li et al. [27] It has been shown that the combination of threshold and morphological function leads to a good segmentation result. Likewise, group segmentation presented by Fiehn et al. [28, 29, 30] are also widely used. Various embryonic cell segmentation techniques including otsu, histogram thresholding and flow markers have also been proposed [31, 32, 33]. Zhang et al. [34] presented HSI (Hue, Saturation and Intensity) color model uses k-means cluster segmentation technique to transform cell images into other models. Ko et al. [35] Segmentation of WBC using Gradient Vector Flow (GVF) snake and stepwise merge rule.

Segmenting blast cells from peripheral blood smears poses a significant challenge due to various factors, including no difference between the objects in the picture and the background, and there is sensitivity to noise. Many existing methods struggle to effectively handle the complexities inherent in such images. Therefore, there is an urgent need to develop innovative and robust techniques specifically tailored for this task.

In this research, an automatic system is developed to diagnose Acute Leukemia, addressing the significant challenges observed in the segmentation and feature selection of blood smear images. The complexity of efficiently segmenting blood cells and selecting pertinent features has hindered accurate diagnosis. To surmount these challenges, the study proposes leveraging flow cytometry test results instead of directly analysing blood cell images. This method aims to increase the accuracy and reliability of Leukemia diagnosis by using detailed cellular information provided by flow cytometry, thus bypassing image analysis problems.

Flow cytometry is a powerful technique used by haematology for the diagnosis and classification of acute Leukemia, among many other applications. This technology allows for the rapid analysis of thousands of cells in a sample of blood or bone marrow. It can differentiate between normal cells and abnormal cells, such as Leukemia blasts, by measuring the light scatter and fluorescence intensity emitted by the cells as they are tagged with fluorescently labelled antibodies and pass individually in a stream through a laser beam. This process generates data that represent various physical and chemical characteristics of the cells, such as size, complexity, and the presence of specific cell surface markers. These data are called Forward Scatter signal, Forward Scatter Pulse Width Signal, Side Scatter Signal, and Side Fluorescence Signal. Here researchers have used such type of dataset for Leukemia diagnosis which can measure the individual patient’s blood cell size and complexity/ granularity. By plotting FSC against SSC, populations of cells can be initially gated or segregated based on size and granularity. Unlike the forward scatter signal (FSC), which mainly provides size estimation based on intensity, FSCW offers additional dimensions of size characterization, particularly useful for identifying cell aggregates or distinguishing between cells of similar sizes but different shapes. SSC provides information on cell complexity, while fluorescence signals identify specific cell types or states based on antigen expression. All types of signals analyzed together to give a comprehensive picture of the cell populations present in a sample. They work together to provide a detailed cellular analysis, enabling the identification, classification, and monitoring of leukemia.

Flow cytometry provides a powerful, high-throughput means of diagnosing acute leukemia by allowing for the rapid and detailed analysis of cell populations in a sample. The discrete data generated capture the complexity and heterogeneity of leukemic cells, facilitating accurate diagnosis and monitoring. The analysis relies on the interpretation of light scatter properties and fluorescence intensity patterns, which reflect the presence and density of specific markers critical to identifying leukemia cells. This approach leverages the comprehensive data provided by flow cytometry to distinguish between healthy individuals and those with leukemia, utilizing both statistical analysis and machine learning techniques for effective classification and diagnosis. For this study, dataset is provided by AIIMS Hospital, New Delhi, 368 patients—half diagnosed with Acute Leukemia and half healthy—each represented by 2000 to 3000 blood cells. The data is split into 70% for training and 30% for testing. The classification methods we use include SVM, KNN, Random Forest, Decision Tree, Logistic Regression and AdaBoost algorithm. The models' performances are evaluated to determine the most effective approach for accurate classification.

Dataset preprocessing and classification methods play crucial roles in accurately predicting outcomes from complex biomedical datasets, such as those involved in diagnosing leukemia using flow cytometry. Flow cytometry datasets, rich in cellular feature data, require thorough preprocessing to correct for issues like noise, irrelevant features, and data imbalances. Preprocessing steps often involve normalization, feature selection or extraction, and data augmentation to improve the model's learning efficiency and prediction accuracy. Classifying leukemia patients into positive and negative categories using flow cytometry data involves several key steps: 1) Data acquisition and preprocessing to prepare the dataset, 2) Choosing and training a suitable classification model with a portion of the preprocessed data, 3) Validating the model using a different subset of the data to fine-tune parameters and avoid overfitting, and 4) Testing the model with unseen data to assess its generalization capability and performance in accurately distinguishing between positive and negative leukemia cases. This pipeline enables the development of predictive models that can significantly aid in the early and accurate diagnosis of leukemia, improving treatment outcomes. In below Fig. 2. mention the steps of classification for leukemia cancer.

4.1 DATASET PREPROSSESSING:

To address class imbalance and enhance model performance, the project integrated Synthetic Minority Oversampling Technique (SMOTE) and Principal Component Analysis (PCA). SMOTE was utilized to artificially augment the minority class through synthetic data generation, effectively balancing the dataset. PCA contributed by reducing the dimensionality of the data, preserving essential information while minimizing computational costs and speeding up model execution. Together, these strategies not only improved the generalization of the dataset but also ensured efficient and effective outcome evaluations, showcasing their adaptability to various data structures and imbalances.

4.2 CLASSIFICATION METHODOLOGY:

K-Nearest Neighbor The algorithm distinguishes leukemia cells and normal blood cells, showing that k-NN is a universal classification tool with good performance. [36].The advantage of k-NN is that there is no need to learn time, it is easy to use and new information can be added without any problems.. By applying the no. of neighbors K = 6,10 and 12 can predict the diagnosis of Leukemia.

Support Vector Machine classification algorithm used to classify Healthy and Blasts blood cells. Support Vector Machines (SVM) employ a clever strategy to tackle nonlinear classification problems [37]. SVM converts the nonlinear separation problem in the input plane into a linear separation problem in the feature space. This linear separation facilitates the creation of a maximum margin hyperplane, optimizing the classification boundary between the given classes. Thus, SVM leverages the power of nonlinear mapping functions to simplify classification tasks in higher-dimensional feature spaces, ensuring robust and accurate predictions. Kernel value rbf, linear and sigmoid are used as a hyperparameter in SVM classification algorithm for prediction.

Random Forest is a popular Integrated learning technique for classification and regression tasks in machine learning. It works by creating multiple decision trees during training and outputting the model (classification) of the cluster. For the classification task, the class with the most “votes” from each tree is selected as the final prediction. Random forests can be large random files [38]. No. of estimators (no. of trees in the forest) is considered as a hyperparameter in RF classification algorithm.

Decision A tree classifier is a supervised machine learning algorithm that uses simple rules determined from data features to predict the probability of a target variable. In this study, a model using a decision tree classifier and its preset was not used, but the model was used for the data set to obtain the results.This approach leverages the intrinsic simplicity and effectiveness of Decision Trees for predictive analysis [39].

Logistic Regression a key algorithm for classification tasks in machine learning, utilizes predictive analysis and probability assumptions to forecast outcomes. Specifically, it constrains predictions within the 0 and 1 range, effectively modeling the probability of events. In our application, we've implemented a penalty parameter with a value of 12 and set the inverse of regularization strength (C) to 0.1, fine-tuning the model's complexity and its ability to generalize to new data [40].

AdaBoost or Adaptive Boosting is a machine learning meta-algorithm designed to enhance the performance of various learning algorithms by focusing on difficult-to-classify instances. It's particularly sensitive to noisy data and outliers. In our application, we've configured AdaBoost with n_estimators = 1000 and random_state = 0, parameters that define the number of sequential models to be trained and the seed for random number generation, respectively, to optimize its performance and ensure reproducibility of results [41].

In this section, we explore the effectiveness of diverse machine learning strategies aimed at boosting the accuracy of leukemia cancer classification. A range of innovative classification techniques was applied to formulate a detection system specifically designed for analyzing flow cytometry test datasets, highlighting the advancements in accurately identifying leukemia. Several machine learning methods are used, such as DT, KNN, RF, AdaBoost, LR and SVM to achieve the highest prediction results. The evaluation results of different machine learning model with evaluation parameters to improve the system’s learning capabilities. Tab.1, Tab.2, Tab.3, Tab.4, Tab.5, Tab.6 shows the highest accuracy achieved during the evaluations and different models DT, KNN, RF, SVM, LR and AdaBoost respectively with different parameters.

Table 1: Analysis and Evaluation of Decision Tree model with different max_depth value

DT: maxdepth	Model Predicted
DT: maxdepth	accuracy	f1score	recall	precision
d=8	0.75	0.37	0.56	0.53
d=12	0.66	0.18	0.14	0.22
d=25	0.53	0.34	0.47	0.29

When a decision tree is too shallow or doesn't have enough branches, it may not capture complex patterns in the data, leading to underfitting. This means the model is too simple to understand the nuances of the dataset, compromising its predictive accuracy.

Table 2: Analysis and Evaluation of KNN model with different n_neighbours value

KNN: n_ neighbours	Model Predicted
KNN: n_ neighbours	accuracy	f1score	recall	precision
n=6	0.67	0.4	0.33	0.5
n=10	0.72	0.46	0.4	0.55
n=12	0.79	0.52	0.47	0.63

The KNN algorithm struggles with high-dimensional data, suffering from the curse of dimensionality, which degrades its performance.

Table 3: Analysis and Evaluation of Random Forest model with different n_estimators value

Random forest: n_estimators	Model Predicted
Random forest: n_estimators	accuracy	f1score	recall	precision
100	0.65	0.43	0.45	0.51
120	0.58	0.4	0.35	0.45
50	0.53	0.38	0.37	0.41

Imbalanced data significantly impacts the classification performance of algorithms like random forests, compromising their accuracy.

Table 4: Analysis and Evaluation of SVM model with different kernel value

SVM: kernel	Model Predicted
SVM: kernel	accuracy	f1score	recall	precision
rbf	0.92	0.8	0.67	0.98
sigmoid	0.8	0.72	0.54	0.87
linear	0.83	0.75	0.5	0.91

High-dimensional data can be managed well with models capable of handling small datasets and modeling non-linear decision boundaries, offering versatility in various applications.

Table 5: Analysis and Evaluation of Logistic Regression model with different regularization value

Logistic Regression:c	Model Predicted
Logistic Regression:c	accuracy	f1score	recall	precision
0.1	0.77	0.48	0.52	0.55
1	0.75	0.52	0.46	0.59
10	0.65	0.4	0.38	0.47

Logistic regression may struggle with small sample sizes, cannot predict continuous outcomes, and is unsuitable for non-linear problems due to its linear decision surface, making it less effective for complex real-world data that's rarely linearly separable.

Table 6: Analysis and Evaluation of AdaBoost model with different n_estimators value

AdaBoost: n_estimators	Model Predicted
AdaBoost: n_estimators	accuracy	f1score	recall	precision
30	0.6	0.34	0.36	0.57
50	0.61	0.42	0.42	0.54
70	0.65	0.55	0.49	0.67

AdaBoost requires high-quality data and is sensitive to outliers and noise, necessitating data cleaning before application, in contrast to Random Forest.

Below Fig.4 shows the performance of different machine learning models with different parameter values on different evaluation criteria.

Table 7, listing machine learning models alongside their best accuracy results based on different evaluation criteria. This table effectively showcases the strengths and weaknesses of each model according to different evaluation criteria. This table not only compares the models across various key performance indicators but also clearly highlights the SVM (Support Vector Machine) model as achieving the best accuracy, making it a potentially more suitable choice for the specific dataset or problem at hand. The highlighted SVM row draws attention to its superior performance metrics across accuracy, precision, recall, F1-score indicating its effectiveness in the evaluated task compared to the other models.

Table 7: Comparative results of machine learning models with different hyperparameter.

Machine Learning Algorithm	Hyperparameter	Model Predicted Results
Machine Learning Algorithm	Hyperparameter	Accuracy	F1 score	Recall	Precision
KNN	n_neighbours=12	79%	52%	47%	63%
RF	n_estimators=100	65%	43%	45%	51%
DT	max_depth=8	75%	37%	56%	53%
SVM	kernel= rbf	92%	80%	67%	98%
Logistic Regression	regularization =0.1	77%	48%	52%	55%
AdaBoost	n_estimators=70	65%	55%	49%	67%

In this study, we introduce a novel methodology for early leukemia detection employing flow cytometry test datasets, which detail the forward scatter and side scatter values of blood cells from patients. Our research conducts a thorough comparative analysis of various machine learning algorithms to identify the most effective model for predicting leukemia presence accurately. Among the evaluated algorithms, the Support Vector Machine (SVM) Classification model emerged as the top performer, demonstrating remarkable efficacy with a 92% accuracy rate, 98% precision, 67% recall, and an 80% F1-score. These metrics not only showcase the SVM model's superior predictive capabilities but also underscore its potential as a reliable diagnostic tool in medical practice. Looking ahead, we propose further exploration into the integration of machine learning algorithms, including SVM, to potentially replace or enhance current convolutional neural network (CNN) classification methods within the realms of medical diagnostics and bioinformatics. This shift could herald a significant transformation in how medical data is analyzed and interpreted for disease detection and patient care.

Conflicts of Interest:

The authors declare that they have no conflicts of interest to report regarding the present study.

Ethics Declaration

Not Applicable.

Funding Statement:

The authors received no specific funding for this study.

Author Contribution

Gupta.R. wrote the main manuscript text and ResearchKapadia.Viral guide

Acknowledgement:

The authors would like to thank the Editor and the anonymous reviewers for their insightful comments and constructive suggestions that certainly improved the quality of this paper.

Data Availability

The experimental data that support the findings of this study are available from All India Institute of Medical Sciences Delhi

S. Agaian, M. Madhukar, and A. T. Chronopoulos, “Automated screening system for acute myelogenous leukemia detection in blood microscopic images,” IEEE Systems journal, vol. 8, no. 3, pp. 995–1004, 2014.
S. Shafique and S. Tehsin, “Acute lymphoblastic leukemia detection and classification of its subtypes using pretrained deep convolutional neural networks,” Technology in Cancer Research & Treatment, vol. 17, Article ID 1533033818802789,2018.
R. D. Labati, V. Piuri, and F. Scotti, “All-IDB: the acute lymphoblastic leukemia image database for image processing,” in Proceedings of the 2011 18th IEEE International Conference on Image Processing, pp. 2045–2048, IEEE, Brussels, Belgium, Sept. 2011.
T. Tran, O.-H. Kwon, K.-R. Kwon, S.-H. Lee, and K.-W. Kang, “Blood cell images segmentation using deep learning semantic segmentation,” in Proceedings of the 2018 IEEE International Conference on Electronics and Communication Engineering (ICECE), pp. 13–16, IEEE, Xi’an, China, Dec. 2018.
F. Xing and L. Yang, “Robust nucleus/cell detection and segmentation in digital pathology and microscopy images: a comprehensive review,” IEEE Reviews in Biomedical Engineering, vol. 9, pp. 234–263, 2016.
J. Wen, Y. Xu, Z. Li, Z. Ma, and Y. Xu, “Inter-class sparsity based discriminative least square regression,” Neural Networks, vol. 102, pp. 36–47, 2018.
Kokeb Dese, Hakkins Raj, Gelan Ayana, Tilahun Yemane, Accurate Machine- Learning Based classification of Leukaemia form Blood Smear Images,2021 Elsevier.
Supriya Mandal, Vani Daivajna, Rajagopalan Machine Learning based System for Automatic Detection of Leukaemia Cancer Cell,2019 ,IEEE.
M.Madhukar, S. Agaian,A. T. Chronopoulos “New decision support tool for acute lymphoblastic leukemia classification”. In Image Processing: Algorithms and Systems X and Parallel Processing for Imaging Applications II , International Society for Optics and Photonics.2012, Vol. 8295, p. 829518.
E. Fathi, M. J. Rezaee, R. Tavakkoli-Moghaddam, A. Alizadeh, and A. Montazer, “Design of an integrated model for diagnosis and classification of pediatric acute leukemia using machine learning,” Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine, vol. 234, no. 10, pp. 1051–1069, 2020.
S. S. Al-jaboriy, N. N. A. Sjarif, S. Chuprat, W. M. Abduallah, and W. M. Abduallah, “Acute lymphoblastic leukemia segmentation using local pixel information,” Pattern Recognition Letters, vol. 125, pp. 85–90, 2019. View at: Publisher Site | Google Scholar
Francis E, Mashor M, Hassan R. 2011. Screening of bone marrow slide images for leukemia using multilayer perceptron (MLP). Paper presented at IEEE symposium on industrial electronics and applications (ISIEA2011).

F. E. Al‐Tahhan, M. E. Fares, A. A. Sakr, and D. A. Aladle, “Accurate automatic detection of acute lymphatic leukemia using a refined simple classification,” Microscopy Research & Technique, vol. 83, no. 10, pp. 1178–1189, 2020.
Mohapatra, S., Patra, D., & Satpathy, S. (2014). An ensemble classifier system for early diagnosis of acute lymphoblastic leukemia in blood microscopic images. Neural Computing and Applications, 24, 1887–1904.

Dumyan, S., & Gupta, A. (2017). An enhanced technique for lymphoblastic cancer detection using artificial neural network. International Journal of Advanced Research in Computer Science and Electronics Engineering (IJARCSEE), 6, 8.
Karthikeyan, T., & Poornima, N. (2017). Microscopic image segmentation using fuzzy C means for leukemia diagnosis. International Journal of Advanced Research in Science, Engineering and Technology, 4.
Saif S. Al-, Nilam Nur Amir Sjarif , Suriayati Chuprat , Wafaa Mustafa Abduallah , Acute Lymphoblastic Leukemia Segmentation Using Local Pixel Information, Pattern Recognition Letters (2019), doi:https://doi.org/10.1016/j.patrec.2019.03.024R. M. Haralick and L.
G. Shapiro, “Image segmentation techniques,” in Applications of Artifical Intelligence II, 1985, vol. 548, pp. 2–10.
S. Yuheng and Y. Hao, “Image Segmentation Algorithms Overview,” ArXiv Prepr.

ArXiv170702051, 2017.

Mohammed, M. A., Ghani, M. K. A., Hamed, R. I., Abdullah, M. K., and Ibrahim, D.A., Automatic segmentation and automatic seed point selection of nasopharyngeal carcinoma from microscopy images using region growing based approach. J. Comput. Sci. 20:61–69, 2017.
A.A. El-Nasser, M. Shaheen, H. El-Deeb, Enhanced leukaemia cancer classifier algorithm, 2014 Science and Information Conference, 2014.
Saraswat M., and Arya, K. V., Automated microscopic image analysis for leukocytes identification: A survey. Micron 65:20–33, 2014.
J. Rawat, A. Singh, H. S. Bhadauria, J. Virmani, and J. S. Devgun, “Computer assisted classification framework for prediction of acute lymphoblastic and acute myeloblastic Leukemia,” Biocybern. Biomed. Eng., vol. 37, no. 4, pp. 637–654, 2017.
J. Rawat, A. Singh, and H. S. Bhadauria, “An approach for leukocytes nuclei segmentation based on image fusion,” in Signal Processing and Information Technology (ISSPIT), 2014 IEEE International Symposium on, 2014, pp. 000456– 000461.
S. Mohapatra, D. Patra, S. Satpathi, Image analysis of blood microscopic images for acute leukaemia detection, 2010 International Conference on Industrial Electronics, Control and Robotics, 2010.
V. Singhal, P. Singh, Local Binary Pattern for automatic detection of Acute Lymphoblastic Leukaemia, 2014 Twentieth National Conference on Communications (NCC), 2014.
Y. Li, R. Zhu, L. Mi, Y. Cao, and D. Yao, “Segmentation of white blood cell from acute Lymphoblastic Leukemia images using dual-threshold method,” Comput. Math. Methods Med., vol. 2016, 2016.
A. S. Negm, O. A. Hassan, and A. H. Kandil, “A decision support system for Acute Leukaemia classification based on digital microscopic images,” Alex. Eng. J., 2017.
S. Agaian, M. Madhukar, and A. T. Chronopoulos, “Automated screening system for acute myelogenous Leukemia detection in blood microscopic images,” IEEE Syst. J., vol. 8, no. 3, pp. 995–1004, 2014.
D. Goutam and S. Sailaja, “Classification of acute myelogenous Leukemia in blood microscopic images using supervised classifier,” in Engineering and Technology (ICETECH), 2015 IEEE International Conference on, 2015, pp. 1–5.
T. G. Patil and V. B. Raskar, “Automated Leukemia Detection By Using Contour Signature Method,” Int. J. Adv. Found. Res. Comput., vol. 2, no. 6, 2015.
S. Mishra, B. Majhi, P. K. Sa, and L. Sharma, “Gray level co-occurrence matrix and random forest based acute lymphoblastic Leukemia detection,” Biomed. Signal Process. Control, vol. 33, pp. 272–280, 2017.
V. Shankar, M. M. Deshpande, N. Chaitra, and S. Aditi, “Automatic detection of acute lymphoblastic Leukemia using image processing,” in Advances in Computer Applications (ICACA), IEEE International Conference on, 2016, pp. 186–189.
C. Zhang , X. Xiao , X. Li , et al. , White blood cell segmentation by color-s- pace-based k-means clustering, Sensors 14 (9) (2014) 16128–16147
B.C. Ko , J.W. Gim , J.Y. Nam , Automatic white blood cell segmentation using stepwise merging rules and gradient vector flow snake, Micron 42 (7) (2011) 695–705.
Endah Purwanti and Evelyn Calista, Detection of acute lymphocyte leukemia using knearest neighbor algorithm based on shape and histogram features, International Conference on Physical Instrumentation and Advanced Materials ,2017.
Morteza Moradi Amin, Saeed Kermani, Ardeshir Talebi, Recognition of Acute Lymphoblastic Leukaemia Cells in Microscopic Images Using K-Means Clustering and Support Vector Machine Classifier, 2015, Journal of Medical Signals & Sensors.
Pouria Mirmohammadi ,Marjan Ameri ,Ahmad Shalbaf, Recognition of acute lymphoblastic leukemia and lymphocytes cell subtypes in microscopic images using random forest classifier, Australasian College of Physical Scientists and Engineers in Medicine, 2021.
Khaled A. S. Abu Daqqa, Ashraf Y. A. Maghari, Wael F. M. . Al Sarraj, Prediction and Diagnosis of Leukemia Using Classification Algorithms,8th International Conference on Information Technology (ICIT),2017 IEEE.
Xiaobo Zhaou, Kuang-Yu Liu, Stephen T.C. Wong, Cancer classification and prediction using logistic regression with Bayesian gene selection, Journal of Biomedical Informatics 37 (2004) 249–259 @ Elsevier Inc.
M. A. Hossain, M. I. Sabik, Md. M. Rahman, S. N. Sakiba , A. K. M. Muzahidul Islam .S. Shatabda , S. Islam, A.Ahmed, An Effective Leukemia Prediction Technique Using Supervised Machine Learning Classification Algorithm, Proceedings of International Conference on Trends in Computational and Cognitive Engineering, Advances in Intelligent Systems and Computing,2021.
Abdul Karim, Azhari Azhari, Mobeen Shahroz, Samir Brahim Belhaouri and Khabib Mustofa, LDSVM: Leukemia Cancer Classification Using Machine Learning, Computers, Materials & Continua · December 2021.
R.N. Agrawal,V.Kapadia, Systematic Review of Diagnosis and Classification Techniques of Acute Lymphoblastic Leukaemia and Acute Myeloid Leukaemia Cells, Indian Journal of Natural Sciences, Vol.14 / Issue 82 / Feb / 2024.

No competing interests reported.

Patient-Centric Diagnosis of Acute Leukemia: A Machine Learning Approach Utilizing Flow Cytometry Data

Status:

Version 1

Abstract

Figures

1. INTRODUCTION

2. RELATED WORK DONE

3. DATASET OVERVIEW AND SPECIFICATIONS

4. MATERIALS AND METHOD

4.1 DATASET PREPROSSESSING:

4.2 CLASSIFICATION METHODOLOGY:

5. PERFORMANCE EVALUATION AND RESULT

6. CONCLUSION AND FEATURE WORK

Declarations