AN EFFICIENT ALGORITHM FOR ACUTE LYMPHOBLASTIC LEUKEMIA QUANTIFICATION IN BLOOD CELL IMAGES USING SUPPORT VECTOR MACHINES ALGORITHM

. As the capabilities of information technologies develop, the effectiveness of using the capabilities of CNN convolutional neural networks, an algorithm with a high level of accuracy, and early prediction of various diseases in blood cell images is high. This study proposes an effective method for early detection of acute lymphoblastic leukemia in blood cell images using Support Vector Machines (SVM) algorithm. In this method, a pattern of cells is recognized and used to identify cell markers specific to leukemia. This algorithm is used to match leukemia to a single marker in the cell image. Comparing SNN convolutional neural network algorithms with random forest (RF), Bayesian classifier, Support Vector Machines (SVM) and K nearest neighbor (KNN) algorithms, the results obtained by Support Vector Machines (SVM) were found to be 90.9% efficient.


I. INTRODUCTION
Today, there are several optimal algorithms for processing and analyzing blood cell images, which are considered to be an effective solution for early diagnosis of all blood-related diseases.For example, this deficiency causes anemia and even leukemia.In pregnant women, such a change is 90%, and in children aged 0-5 years and above, this indicator is 80%.The composition of the bone cells consists mainly of three types of cells: erythrocytes, leukocytes and thrombocytes [1][2][3][4].There are five subclasses of leukocytes, which are monocytes, lymphocytes, basophils, eosinophils, and neutrophils.Multiclass classification is considered to be the best approach for the diagnosis and precise identification of leukocytes and their subclasses because it can be used to quickly identify each class [5].
Various image features such as edges, geometric, statistical and statistical features as well as histogram of gradients (HOG) are used to classify images.Preprocessing, which includes noise reduction, contrast control, and image sharpening, is the first step in image classification.Various techniques are used to enhance microscopic images.The enhanced image is then further processed to separate WBCs using different segmentation methods [6][7][8].
In several disciplines, including medical diagnostic systems, significant patterns have been extracted for prediction tasks using machine learning-based networks [9].If the disease is detected early and treated promptly, the death rate can be reduced.This makes things hopeless and it takes a long time for a hematologist to physically find the disease.To overcome these problems, computer-aided methods for leukemia detection are very effective, fast and accurate [10].However, the computer-assisted approach still faces a number of challenges, obstacles, and research gaps, such as the accuracy of detecting leukemic cancer and its types (acute myelogenous leukemia, acute lymphoblastic leukemia, and multiple myeloma).their further segmentation and categorization.The purpose of this study is to identify the subtype of leukemia and to diagnose the disease.

II. METHODOLOGY
Leukocyte segmentation and segmentation, which currently play an active role in medical hematology for the diagnosis of liver disease, enable the relevance of leukocyte segmentation and segmentation for the classification and segmentation of leukemia.
In [11,12]  A threshold and mathematical morphology [13] were used to distinguish the cell nucleus.Morphology is a mathematical method that differentiates between white blood cells (WBC), red blood cells (RBCs), and platelets.This [14] technique applies addition and subtraction to blood smear images.The image was segmented into background and foreground using threshold segmentation, and then the best threshold value was selected for WBC segmentation.To classify leukocytes, geometric features were extracted and SVM classifier was applied [15].
Several strategies have been presented to address the problem of blood cell matching.These algorithms separate cells by erosion and growth, which preserve the shape or merge dull spots with dividing lines [16 ,17].As part of this research, we also propose a cell separation algorithm.a conic curve separating overlapping parts [18].

III. THE PROPOSED METHOD
1.The blood sample applied to the proposed result architecture involves several steps. 1 Input photos are first collected 2 These photos often have different sizes and resolutions.3 They are useful for any experiment and analysis due to their uniform size.4 these photos are then scaled and color filtered Random forest, Naive Bayes classifier, support vector machine and logistic regression are used for classification.Figure 1 illustrates each step of the proposed methodology.In the preprocessing stage, various operations such as resizing, denoising, and contrast adjustment are performed, and then features are extracted.After that, dimension reduction is applied for PCA and finally classification is done using classifiers.In the final step, a multi-class classifier is used for classification.
Pseudocode for an algorithm for quantifying leukemia from blood cell images.
Stage I: Image preprocessing algorithms Step 1 Blood cell imaging Input: Raw digital images of blood cells.
Step 2. Resize the image to 256x256 pixels Often there is some local distortion in the image, which is caused by light diffraction, imperfect optical systems, or defocusing.This leads to the need to make local changes to the image.In other words, this flexible approach allows you to select areas of information in the image and process them accordingly.The methods of adaptive change of local contrast meet the specified requirements [3].The following symbols are used for this: , ( , ) xy initial image and its element according to coordinate; The disadvantage of the image is that the image is low-contrast, which makes it difficult to analyze the image and causes an overestimate in the calculations.Therefore, first of all, the operation of stretching the image histogram to the maximum permissible range is performed.This improves the visual quality of the images.0 and L-1 r there is a gray level M×N An image of pixel size F is considered.When using a fuzzy set for image processing, they are treated as an array of fuzzy singletons.Each element of the array indicates the corresponding value of the gray level corresponding to the pixel according to predefined characteristics such as brightness, clarity, uniformity of the ( , ) As a generalization of this approach, we present the following view of an image in a fuzzy environment.
The image F depicted in an opaque environment has the following form  From the point of view of fuzzy image processing, the question that naturally arises when trying to determine the brightness of a fuzzy pixel can be formulated as follows: "How can we determine the function of gray levels to describe an image in a fuzzy set?" Uncertainty in images is caused by various factors.They decide whether a pixel is "grey" or "brighter" .
Algorithm of linear adaptive enhancement of image contrast: Step 3. Normalization is performed: Step 4. Fuzzification is performed: Step 5. Fuzzification, that is, further refinement is carried out at the limits of refinement: Step 6.
here  denotes the number of elements of the fuzzy set.
In addition, ( ) appearance of the brightness level.However, due to its definition, the fuzzy histogram is not a probability density function.
Step 7. The normalized histogram is determined by the corresponding functions: ). ( , ) Step 8. Contrast is defined by the following expression: where ( , ) C x y according to the expression (1), equal to zero for a place with constant intensity, and 2 () L  equal to 1 for large values of the property of this (1) expression fully meets the requirements for determining the local contrast of a given section.
Step 9. F The degree of ambiguity can be defined by analogy with the Shannon entropy in the form [5][6][7], and this is also improved using relevance functions: We repeat the described procedure for each image element.
The recommended method uses statistical determination of local contrasts, which takes into account such features as texture uniformity, roughness, and granularity.Therefore, this method is recommended for processing pictures containing small details.
Stage II: Support vector machine: Step 1. Input: pre-processed image; Step 2. RPN is started; Step 3.For each proposal, the cell division probability and location coordinates (bounding box) are calculated; Step 4. Features are extracted using RoI Pooling; Step 5. Specify the cell type and location Model optimization using Transfer Learning; Step 6.Using the pre-trained model; Step 7. Freeze the initial layers of the model and retrain the last layers; Step 8. Cells identified and classified.
Stage III: Diagnosis: Step 1. Input: Identified and classified cells; Step 2. Separation of cells by erythrocytes and leukocytes; Step 3. Count the number of cells of each type; Step 4. Diagnosis of Leukemia.3. Release features.features can be extracted from objects in the image.Because blast cells (ROIs) contain a lot of information, including information about their cytoplasm and nucleus, the feature extraction step is very important to identify the type of acute leukemia.The resulting photographs show round cells.For convenience, an intelligent edge detection algorithm is applied to each image during feature detection.A Gaussian filter is used to smooth the image and remove noise.Next, we calculate the intensity changes within the image.A two-layer threshold is then applied to determine the possible boundaries of the image.Hysteresis is then used to trace the contours of the image.Weak edges unrelated to strong edges are suppressed to complete the edge detection process.Figure 3 shows step-by-step effects on sample images.Figure 3a shows a sample image after processing and color filtering; Figure 3b shows the image after the operation.Figure 3c shows the effect after the feature extraction process and Figure 3d shows the effect after the feature description process.

IV. DISCUSSION
Both qualitative and quantitative aspects of the experimental results of the proposed hybrid model classification methodology are presented.Using the obtained data, we tested the proposed strategy.For diagnosis, leukocytes are divided into special segments so that the structure and color of the nuclei are clearly visible.The underlying truth was the opposite of the proposed technique.This algorithm is able to accurately identify and distinguish five types of leukocytes.Three metrics were used to evaluate the proposed segmentation technique: false-positive rate (FPR), false-negative rate (FNR), and F-measure.Table 1 shows the results obtained from the proposed model.Table 1 focuses on the performance of different classifiers used in the study.Table 1, which shows the comparative analysis, describes the precision, accuracy, recall and F-score values for the classifiers.Support vector machine obtained the best performance metric results compared to other classifiers.Figure 4 depicts the comparison graph used for the analysis.This figure focuses on the performance of the classifier in terms of performance metrics identified in the study.As you can see from the table, comparing different algorithms, the difference between them is shown in the diagram above: Random forest, 90.5% accuracy, Support vector machine, 90.9%, Nearest neighbors, 90.07%, Naive Bayes, 89.90%, and the exact error and growth percentage are clearly shown.The proposed algorithm was the highest Support vector machine 90.9%.

V. CONCLUSION
The algorithm proposed above is distinguished from other algorithms by its highest level of effective accuracy.By further improving the Support vector machine 90.9% algorithm, it will be possible to obtain more effective results from it in the future.The use of this method in medicine is more effective and faster than other algorithms, and it helps patients with early-detected diseases.provides an opportunity to prevent diseases.Leukemia is a type of blood cancer that often affects children and adults.The type of cancer and the extent to which it has spread throughout the body affect the treatment of leukemia.The disease should be detected as soon as possible so that the patient can be given proper care and treatment.This study presents a new strategy for fully automatic detection and classification of leukocytes using microscopic images.The aim of this work is to provide an automated technique to support medical activities in the detection of acute lymphocytic leukemia (ALL).
developed a number of mathematical operations and techniques to reduce noise and increase image clarity.Procedures are used to enhance gamma structure and contrast.The accuracy of the calculation is widely segmented.Leukemia localization and f-region extraction were the first two steps of the two-step image segmentation technique.There are three more steps in each technique.Localization thresholding, three-phase filtering, neighborhood detection, and cell extraction are examples of substeps.Cytoplasm, nucleus, and nucleus localization are also extracted.
define the level of the pixel belonging to the set according to the characteristics of the image.

F
functions correspond to many components of an image.Batch image processing techniques provide a flexible mathematical framework for dealing with "quality" features such as image contrast in situations of ambiguity and blurring that are common in digital images.
the discrete intensity of the image, and be the corresponding values of the histogram.The -moment W of n relative to the average value of the brightness of local elements (

Fig. 1 .
Fig. 1.Proposed basis.1.Data collection.All-IDB1 and own photographs from pathology were used to obtain microscopic images of white blood cells for the proposed technique.There are 1208 photos in this dataset.A total of 549 of these photos are good, 659 are dangerous.There are approximately 93,000 blood components.Lymphocytes are labeled in this data set.There are 5510 lymphoblasts among the components.Figure 2 depicts sample images from the dataset under different conditions.These images are used to extract features and obtain a final classification.

Fig. 2 .
Fig. 2. Sample images from the dataset.2. Image preprocessing.Redundancy can be very useful in image processing.Adjacent pixels representing the same physical object have similar or identical brightness values.If a deformed pixel is located in the image, it can be restored by taking the average value of the values of the immediately surrounding pixels.One way to classify image preprocessing methods is by the size of the pixel neighborhood used to calculate the brightness of a new pixel.In the pre-processing stage, we removed the background, cut out excess blood

Fig. 4 .
Fig. 4. Сomparative diagram Precision Errors Graphical representation of the results of comparing Random forest Support vector machine Nearest neighbors with Naïve Bayes models, taking into account the Growth feature.

Table 1 .
Precision Errors Results of comparison of Random forest Support vector machine Nearest neighbors with Naïve Bayes models taking into account the growth feature.