Efficient computer‐aided diagnosis technique for leukaemia cancer detection
2020; Institution of Engineering and Technology; Volume: 14; Issue: 17 Linguagem: Inglês
10.1049/iet-ipr.2020.0978
ISSN1751-9667
Autores Tópico(s)Smart Agriculture and AI
ResumoIET Image ProcessingVolume 14, Issue 17 p. 4435-4440 Research ArticleFree Access Efficient computer-aided diagnosis technique for leukaemia cancer detection Alan Anwer Abdulla, Corresponding Author Alan Anwer Abdulla alan.abdulla@univsul.edu.iq Department of Information Technology, University of Sulaimani, Sulaimani, IraqSearch for more papers by this author Alan Anwer Abdulla, Corresponding Author Alan Anwer Abdulla alan.abdulla@univsul.edu.iq Department of Information Technology, University of Sulaimani, Sulaimani, IraqSearch for more papers by this author First published: 12 February 2021 https://doi.org/10.1049/iet-ipr.2020.0978AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinkedInRedditWechat Abstract Computer-aided diagnosis (CAD) is a common tool for the detection of diseases, particularly different types of cancers, based on medical images. Digital image processing thus plays a significant role in the processing and analysis of medical images for diseases identification and detection purposes. In this study, an efficient CAD system for the acute lymphoblastic leukaemia (ALL) detection is proposed. The proposed approach entails two phases. In the first phase, the white blood cells (WBCs) are segmented from the microscopic blood image. The second phase involves extracting important features, such as shape and texture features from the segmented cells. Eventually, on the extracted features, Naïve Bayes and k-nearest neighbour classifier techniques are implemented to identify the segmented cells into normal and abnormal cells. The performance of the proposed approach has been assessed through comprehensive experiments carried out on the well-known ALL-IDB data set of microscopic blood images. The experimental results demonstrate the superior performance of the proposed approach over the state-of-the-art in terms of accuracy rate in which achieved 98.7%. 1 Introduction Digital image processing is a process of distribution and/or decomposition of signals in which the input of the process is an image and the output might be data, image, or features related to the input image [1]. With the rapid advancement of digital communication technologies and exponential evolution in the use of the Internet, applications for digital image processing have increased significantly in various fields, such as pattern recognition, biometrics, multimedia security, medical image processing, etc. Digital image processing techniques are therefore facilitated in order to accomplish faster and more accurate information. Since health is of major importance, the healthcare industry has been looking for advanced medical procedures and treatment practices that would incorporate with technology in terms of computation and progression in hardware resources [2]. For this purpose, digital image processing techniques could help clinicians to detect and identify diseases, and consequently, to diagnose and treat them. Digital image processing techniques have therefore become increasingly important for the processing of medical images and the detection of abnormalities from those images [2]. Accordingly, computer-aided diagnosis (CAD) systems based on image processing become an interesting topic in medical image processing research area and is a computer-based system that helps medical professionals in diagnosing of diseases from medical images such as X-ray, magnetic resonance imaging (MRI), computed tomography (CT), ultrasound, and microscopic images [3]. Medical image processing advances include compression of the medical image, de-noising medical image, the fusion of medical images, tumour detection, lung cancer detection, breast cancer detection, blood cancer (leukaemia) detection, dental caries detection, etc. This paper is aimed to detect the type of blood cancer (leukaemia) known as acute lymphoblastic leukaemia (ALL) from microscopic blood images. Leukaemia is a Greek term, literally means white blood [4]. It is a type of cancer that affects the blood, lymphatic system, and bone marrow. Leukaemia can be classified according to the growth rate into acute (which spreads rapidly) and chronic (which spreads slowly). Additionally, leukaemia can be divided into lymphoid (which induces lymphocytes) and myeloid (which gives rise to white blood cells (WBCs), red blood cells (RBCs), and platelets) depending on the type of stem cell affection. Consequently, leukaemia can be categorised into four types: ALL, acute myelogenous leukaemia (AML), chronic lymphoblastic leukaemia (CLL), and chronic myelogenous leukaemia (CML) [5, 6]. In general, blood consists of the following components: WBCs, RBCs, platelets, and plasma [4, 7]. Leukaemia influences WBCs, it is a disease of the bone marrow that leads to producing abnormal and immature WBCs, often referred to as blast cells or leukaemia cells. These cells do not function properly which make the body susceptible to various infections, and as they overgrow, they prevent the formation of other normal blood cells resulting in anaemia, bleeding, bruising, etc. [8]. Fig. 1 shows two blood samples of the person affected by ALL and of the normal person. Fig. 1Open in figure viewerPowerPoint Microscopic image of a blood sample (a) ALL case, (b) Normal case This paper proposes a new approach to detect ALL-leukaemia cells and investigating the impact of different features, namely texture, and shape features, on the classification step. The main objective of this paper is increasing the accuracy rate of detecting ALL-leukaemia cells and this leads to reduce the error rate to treat the patients incorrectly in comparison with traditional systems. The main contributions of the proposed approach segment the WBCs properly and fuse the most significant features that extracted from the segmented cells. The rest of the paper is organised as follows: The literature review is discussed in Section 2. The proposed CAD approach is detailed in Section 3. The experimental results are presented in Section 4. Finally, the conclusion about this work is drawn in Section 5. 2 Literature review Through significant advances in digital technology and a substantial increase in computer power, the processing of medical images; in particular, microscopic images of the blood, plays an important role in the early detection of leukaemia and its subtypes. This section reviews the most important and related existing works on detecting ALL-leukaemia cells based on digital image processing. The main competition in this research area is to increase the accuracy rate for the detection of leukaemia cells. This can be achieved by either enhancing the tested microscopic images to accurately segment the WBCs and/or extracting the most important features that have an impact on the accuracy rate. Scotti et al. were first to investigate this field of study in 2004 [9], who proposed a system that can differentiate WBCs from other blood components such as RBCs, platelets, and plasma, in a microscopic image. The system also uses a neural classification technique to classify WBCs into subtypes. The WBC subtypes include eosinophils, neutrophils, basophils, lymphocytes, and monocytes. Low-pass and bandpass filters were used in this system to decrease noise, which was followed by the use of contrast enhancement to minimise non-uniformity of the background. In addition, using membrane detection and nucleus detection, cropped/segmented images were produced and an opening operator technique was used to extract the remaining components from the segmented images, such as RBCs and platelets. Eventually, shape-based features were extracted from the cropped images to be used for classification purposes using the k-nearest neighbour (kNN). In 2005, Scotti suggested another system to classify ALL cells from microscopic images of blood using an automated morphological technique [10]. The system can differentiate leukocytes from other blood components and used an edge detection technique to select lymphocytes from images through membrane selection of lymphocytes. In addition, using Otsu's threshold technique the nucleus and cytoplasm are segmented from WBCs. Furthermore, significant features such as shape, mean, and standard deviation were extracted from the segmented WBCs. By using the kNN classifier technique, the extracted features were classified. To better enhance the microscopic image and to better segment the WBCs compared to the previous works discussed above, Scotti et al. introduced another leukaemia detection approach in 2006 that involves the following steps: pre-processing for noise reduction, segmentation using fuzzy k-means clustering and the Zack algorithm [11]. Such enhancement results in more accurate segmentation of WBCs. To achieve a better accuracy rate, Reta et al., in 2010, developed a system to classify both types of acute leukaemia, namely AML and ALL. The leukocyte, its nucleus, and cytoplasm are segmented by taking advantage of texture and colour information of images pixels. In addition, the system extracts the region of the cell by designing colour and texture information depending on the Markov random field (MRF). The system was applied on bone marrow images and for the segmentation process, it achieved an accuracy rate of 95%. Furthermore, the system extracts certain useful features such as statistical, geometrical, size ratio, and texture features from the nucleus, cytoplasm, and the entire cell. Finally, decision tree, regression function, meta-classifier, and kNN classifier techniques were used for classification purposes [12]. Unlike the previous systems discussed above, which directly extract features from the RGB version of microscopic images, new approaches were proposed that rely on a grey version of the microscopic images. In 2013, Minal et al. introduced a new approach to segment and classify WBCs from microscopic blood images by converting the RGB image to grey image and then certain enhancement techniques, such as histogram equalisation and linear contrast stretching are applied to enhance the microscopic images [13]. In addition, Otsu's thresholding technique is applied for segmenting WBCs from the microscopic images. To remove the remaining unwanted objects, the morphological opening operator has been used. Finally, features such as area, perimeter, and circularity were extracted from the segmented WBCs to classify the lymphocyte as normal or abnormal using kNN classifier technique. This approach achieved an accuracy rate of 93%. All the previous algorithms discussed above have a limitation of treating with the overlapped WBCs. To overcome this limitation, in 2015, Himali et al., used shape-based features, such as area, perimeter, convex hull, roundness, major axis, minor axis, and standard deviation to detect different types of WBCs (such as eosinophil, basophil, neutrophil, monocyte, and lymphocyte), to detect the number of overlapping and non-overlapping cells, and to count the blood cells. In the first step, the RGB image is transformed into a grey image, and then Otsu's thresholding technique was used to convert the image into a binary image, followed by applying certain morphological operations like opening and hole filling. The accuracy rate of 95.8% was obtained in this proposed system [14]. Continuously and regarding the grey image, Ashwini et al. proposed a technique in 2016 to detect leukemic cells from microscopic blood images [15]. This technique first transforms the RGB microscopic image into a grey image and then applies some enhancement techniques such as histogram equalisation and filtering technique to the grey image. In addition, k-mean clustering technique has been used to segment the nucleus of the cells. Secondly, the technique extracts feature, such as the grey level difference method (GLDM) and grey level co-occurrence matrix (GLCM) from the segmented nucleus. Finally, SVM classifier has been used to classify input images as normal or abnormal. To further increase the accuracy rate and differ from the previous systems’ strategy, discussed above, the authors focused on the new colour space known as L*a*b colour space. In 2018, Kumar et al. proposed a new scheme for the detection of leukaemia cells by first applying morphological cleaning to enhance the tested image of the blood. The colour k-mean clustering is then applied to the L*a*b colour space of the blood image to segment the WBCs from the tested microscopic image [16]. In addition, the bounding box technique is applied on the (a*) component of the L*a*b colour space to crop blast cells from the microscopic blood images. Finally, the cropped images are utilised to extract features, such as colour, textural, geometrical, and statistical features to classify each sub-image as infected or benign based on the kNN and Naïve Base classifier techniques. As the authors reported, this proposed scheme achieved the accuracy rate of 92.8%. Hariprasath et al. proposed another system, in 2019, for the detection of leukemic cells based on the morphology of the cell [17]. Like many previously discussed approaches, some enhancement techniques, such as converting an RGB image to CMYK and histogram equalisation were implemented. Additionally, Zack algorithm was used to segment the image followed by certain morphological operations such as dilation and erosion. To segment, the nucleus of the cell, the Otsu's thresholding is applied on the green channel of the RGB image and is combined with the binary image of the a* component of L*a*b* colour space. Finally, features such as shape, statistical, and textural features are extracted from the segmented nucleus and then the SVM classifier technique was used for classification purposes. As the authors claimed, this technique reached 90% of accuracy. Hegde et al., in 2019, developed an automated method for detection of leukaemia by extracting shape, colour, and texture features from segmented cells [18]. The SVM classifier was used for classification of white blood cells into normal and abnormal and an accuracy rate of 92.8% has been obtained. Recently, certain approaches were introduced in 2020. Kumar et al. developed an algorithm which extracts the image-level features and statistical features from the segmented cells [19]. Then, the selected features are applied to the classifier technique to detect the leukaemia cells. As the authors claimed, this algorithm achieved an accuracy rate of 98%. Janaki proposed a system to identify leukaemia from the microscopic blood images. This scheme entails a pre-processing step using adaptive filtering to remove the noises in the images, segmentation step using multi-module sub-clustering to segment the WBCs from the blood images, feature extraction step in which some features such as correlation, energy, and entropy were extracted from the segmented WBCs, and finally, classification step in which the Gaussian feature convolutional visual recognition was used to classify WBCs into normal or abnormal. This technique was obtained 95.59% of the accuracy rate [20]. Zhana et al. proposed a new algorithm for segmenting WBCs from microscopic blood images based on the thresholding segmentation technique and the authors compared their results with the most commonly used segmentation technique which is known as colour-k-means clustering. As the authors reported, their thresholding-based proposed segmentation technique outperforms the colour-k-means clustering [21]. Continuously, in 2020, another CAD system for recognising the blast cells from bone marrow images in which the leukaemia can be identified was developed by Nikitaev et al. To separate the nucleus of the cells from the input images, two segmentation techniques were used. The first one works on the basis of the histogram analysis, while the second one works on the basis of a watershed. Colour and texture features were extracted from the segmented nucleus, and finally, different classifier techniques such as kNN, SVM, and random forest have been used to classify the WBCs into normal and abnormal. As a result, the accuracy rate for each of the classifiers was reached 89, 85, and 80%, respectively[22]. Bodzas et al. also developed an algorithm, in 2020, for leukaemia identification in which introduces a three-phase filtration algorithm to achieve the best segmentation process [23]. Moreover, sixteen robust features were extracted from the segmented images which significantly increased the capability of the classifiers to recognise leukemic cells in the blood images. Furthermore, the SVM classification technique was used to classify normal and abnormal cells. As the authors reported, their system achieved an accuracy rate of 96.72%. Lately, there are certain leukaemia detection techniques were developed based on deep learning. Cecilia et al. proposed a method for recognising white blood cells from microscopic blood images and classify them as healthy or affected by leukaemia and an accuracy rate of 94.1% was obtained [24]. The presented system has been evaluated using SMC-IDB, the IUMS-IDB, and the ALL-IDB datasets. Maíla et al. also proposed a technique for leukaemia cell detection based on convolutional neural networks (CNNs) architecture which capable of differentiating ALL-leukaemia cell [25]. The experiments were performed using 16 data sets with 2415 images. As the authors claimed, this technique reached an accuracy of 97.18%. Saif et al. introduced a system based on machine learning approach and image processing technique, and hence, the characteristics of blast cells were extracted using four-moment statistical features and artificial neural networks (ANNs) [26]. This system is reached an accuracy rate of 97% using ALL-IDB data set. Apart from leukaemia cancer, there are also some approaches developed for medical image for segmenting and detecting different cancers such as in [27-33]. The rest of this paper focuses on the extension and further refinement of the strategy of using digital image processing to increase the accuracy rate for ALL-leukaemia cell detection. Consequently, the efficiency of the proposed approach is evaluated in terms of accuracy rate. 3 Proposed approach This section presents the detailed of the proposed approach which includes the following steps: the pre-processing step to enhance the microscopic image quality, the segmentation step to separate WBCs from other blood components (to detect ALL-leukaemia the WBCs need to be cropped from the microscopic image of the blood), the extraction step to extract the most affected features, and finally, the classification step that helps the system decide whether the cell is normal or abnormal. The details of the steps are as follows: Pre-processing: Originally the input/tested images are in the form of RGB colour space. In this step, the RGB images are converted to L*a*b* colour space. Segmentation: For identifying any type of leukaemia, WBCs need to be considered; as this type of cancer affects only the WBCs. For this reason, WBCs must be separated from other objects inside the image (such as RBCs and platelets). A segmentation technique, colour-k-means clustering, is applied on L*a*b* colour space images to extract/crop the WBCs from microscopic blood images and separate them from other blood components, namely RBCs and background. This technique segments microscopic blood images into three clusters, the first cluster represents RBCs, the second cluster represents WBCs, and the third cluster represents background (which includes platelet and plasma) as illustrates in Fig. 2. The motivation behind using colour-k-means clustering is that this technique is very robust to segment the image into several clusters based on the similarity between objects of one cluster, and the tested image is segmented into three clusters which are (RBCs, WBCs, and background/platelets). Consequently, the focus is only on the cluster that contains WBCs, and the other two remaining clusters are neglected. Fig. 2Open in figure viewerPowerPoint Colour-k-means clustering-based segmentation technique (a) Original image, (b) First cluster, (c) Second cluster, (d) Third cluster Cropping WBCs: Once the colour-k-means clustering technique is implemented on the L*a*b* colour space, the a* component contains information about the WBCs. Meanwhile, shapes of WBCs within a* component are clearly appeared. Thus, the bounding box technique is applied on the original image based on the a* component to highlight and crop the WBCs as individual images, Figs. 3-5 illustrate the block diagram of the second and third steps of the proposed approach. Fig. 3Open in figure viewerPowerPoint Highlighting WBCs in an original image using bounding box technique Fig. 4Open in figure viewerPowerPoint Cropped WBCs Fig. 5Open in figure viewerPowerPoint Block diagram for segmenting and cropping WBCs Feature extraction: To extract the significant features, two versions of the cropped image, obtained in step 3, must be produced. The first version is the binary image after the thresholding (T < 100) segmentation technique is applied on the green channel of the cropped image and this results in segmenting the nucleus from the WBCs. The second version is the grey image of the cropped image. The main reason behind producing these two versions of images is because the intended features must be extracted from the binary image and some others must be extracted from the grey image. For example, shape feature extraction techniques extract features only from the binary image while texture feature extraction techniques extract features only from the grey image. The following are the features in which extracted in this step of the proposed approach: (i) Shape features include area, perimeter, circularity, radius, major-axis, minor-axis, eccentricity, extent, solidity, orientation, filled area, and rectangularity. (ii) Texture features are summarised in Table 1. Table 1. Extracted texture features Feature extraction technique Name of the feature Tamura coarseness direction grey level difference statistics (GLDS) angular second moment contrast mean entropy grey level run length matrix (GLRLM) SRE LRE grey level non-uniformity run percentage run-length non-uniformity low grey level run emphasis high grey level run emphasis Classification: Eventually, Naïve Bayes and kNN classification techniques have been implemented on the extracted features to segregate normal and abnormal cells. Both techniques were evaluated using the k-fold-cross-validation technique for k = 5. The general block diagram of the proposed approach presents in Fig. 6. Fig. 6Open in figure viewerPowerPoint Block diagram of the proposed approach 4 Experimental results The essential objective of the proposed approach is to classify ALL-leukaemia cells into normal or abnormal. In this section, experiments are performed comprehensively to assess the performance of the proposed approach in terms of confusion matrix measurements, in particular, the accuracy rate. In addition, the proposed approach is compared to the most recent existing approaches. 4.1 Dataset The tested input images are taken from a public and well-known dataset called ALL-IDB consisting of two groups of images. The first group is known as ALL-IDB1, which was designed to test segmentation techniques, and consists of 108 microscopic blood images (49 abnormal images and 59 normal images) [10]. The second group is called ALL-IDB2 which was designed to test the classification techniques and contains 260 cropped WBCs images (50% abnormal cells and 50% normal cells) which were taken from ALL-IDB1 [10]. To evaluate the performance of the proposed approach, the first group was considered. 4.2 Results In the pre-processing step, the RGB images were converted to the L*a*b* colour space because the colour-k-means segmentation technique is commonly applied on the L*a*b* colour space. Since the a* component of the L*a*b* colour space contains more information about WBCs [34], thus this component is used for the next step, which is the cropping of WBCs as individual images. Accordingly, all WBCs are cropped from the original image based on the segmented image, the cluster that holds WBCs, using bounding box technique. As a result, 309 WBCs are cropped from abnormal images and 96 WBCs are cropped from normal images. These individually cropped images are therefore used for extracting features. All the extracted features are used to test the accuracy of the proposed approach based on the two classification techniques which are kNN and Naïve Bayes. To evaluate the impact of each extracted feature on the accuracy rate for the classification of WBCs, all features are examined separately, and then they combined/fused with each other. The results are presented in the following tables using both kNN and Naïve Bayes classifier techniques. From Tables 2 and 3, it is quite obvious that the higher accuracy rate is achieved when all the extracted features are combined/fused for both kNN and Naïve Bayes classifiers which are 98.7 and 95.8%, respectively. In addition, the kNN classifier technique provides higher accuracy rate in comparison with Naïve Bayes. Furthermore, the proposed approach is compared to other existing approaches in [16, 18, 22] and the results are presented in Tables 4 and 5. Table 2. Accuracy rate of the tested features using kNN Features Sensitivity, % Specificity, % Accuracy, % Precision, % F1 score, % FPR, % Tamura 65.1 79.4 72.2 86 74.1 20.5 GLRLM 92.4 97.8 95 98 95.1 2.1 GLDS 98.9 97 97.9 97 97.7 2.9 Tamura + GLRLM 95.1 97.9 96.5 98 96.5 2 Tamura + GLDS 98 98 98 99 98.9 1.9 GLRLM + GLDS 96.1 98.9 97.5 99 97.5 1 Tamura + GLRLM + GLDS 98 98 98 99 98.5 1 shape features 91.4 95.7 93.5 96 93.6 4.2 all features (shape and texture) 98 99 98.7 99 98.5 1 Table 3. Accuracy rate of the tested features using Naïve Bayes Features Sensitivity, % Specificity, % Accuracy, % Precision, % F1-score, % FPR, % Tamura 75 81.8 78.4 84 79.2 18.1 GLRLM 69.4 86.9 78.1 91 78.4 13 GLDS 86.2 87.7 86.9 88 87 12 Tamura + GLRLM 82.1 90.9 86.5 92 86.7 9 Tamura + GLDS 92 92.9 92.4 93 92.4 7 GLRLM + GLDS 84.8 94.3 89.5 95 89.6 5.6 Tamura + GLRLM + GLDS 90.5 95.7 93.1 96 96.4 4.2 shape features 89.6 94.6 92.1 95 91.2 5.3 all features (shape and texture) 94.1 95.9 95.8 96 95 4 Table 4. Accuracy rate of the tested approaches using kNN Approaches Sensitivity, % Specificity, % Accuracy, % Precision, % F1-score, % FPR, % proposed 98.2 99 98.7 99 98.5 1 Kumar et al. [16] 92.6 97.1 95.5 97 94.2 3 Hegde et al. [18] 85.5 100 92.8 99 92.3 0 Nikitaeva et al. [22] 88.3 90 89.3 91 89.2 1 Table 5. Accuracy rate of the tested approaches using Naïve Bayes Approaches Sensitivity, % Specificity, % Accuracy, % Precision, % F1-score, % FPR, % proposed 94.4 95.3 95.8 96 95.7 0.5 Kumar et al. [16] 86.3 94.8 93.1 95.1 90.2 1 Hegde et al. [18] 86.8 93.8 90.3 95 90.8 1 Nikitaeva et al. [22] 86.5 89.4 87.6 90 88.2 2 Table 4 demonstrates that the proposed approach achieved higher accuracy than in [16, 18, 22] based on the kNN classifier technique. Table 5 also demonstrates that the proposed approach obtained a higher accuracy rate than in [16, 18, 22] using the Naïve Bayes classifier. More experiments are also conducted to compare the proposed approach to the systems that are based on deep learning such as in [24-26], see Table 6. Table 6. Comparison results between the proposed approach and approaches based on deep learning Approaches Sensitivity, % Specificity, % Accuracy, % Precision, % F1-score, % FPR, % proposed 98.2 99 98.7 99 98.5 1 Cecilia et al. [24] 94 95.1 94.1 95 94.4 2 Claro et al. [25] 96.9 97.1 97.18 96 95.9 1 Al-jaboriy et al. [26] 96.8 97.2 97 97 96.8 1 Based on the results in Table 6, the proposed approach also exceeds the accuracy rate in [24-26]. 5 Conclusion The advancement of the application of medical image processing in the field of healthcare has led to an improvement in the accuracy of early detection of the disease since the detection of a disease manually is highly costly, time-consuming, and require professional staff. Meanwhile, medical images processing plays an important role in the early identification and detection of various diseases, in particular cancers. The computer-based disease detection systems refer to as CAD system. CAD systems can extract target diseases accurately at higher speeds and lower costs. This paper concentrated on early detection of blood cancer, known as leukaemia. The experimental results demonstrate that the proposed approach earned an accuracy rate of 98.7% and it exceeds the existing approaches. 6 References 1Khedaskar S.V., Rokade M.A., Patil B.R. et al.: ‘A survey of image processing and identification techniques’, VIVA-Tech. Int. J. Res. Innov., 2018, 1, (1), pp. 1– 10 2Kumar A.G., Murthy N.V.: ‘Analysis of medical image processing and its applications in healthcare industry’, Int. J. Comput. Technol. Appl., 2014, 5, pp. 851– 860 3Arimura H., Magome T., Yamashita Y. et al.: ‘Computer-aided diagnosis systems for brain diseases in magnetic resonance images’, Algorithms, 2009, 2, pp. 925– 952 4Kekre H.B., Archana B., Galiyal H.R.: ‘Segmentation of blast using vector quantization technique’, Int. J. Comput. Appl., 2013, 27, pp. 20– 23 5ElDahshan K.A., Youssef M.I., Masameer E.H. et al.: ‘An efficient implementation of acute lymphoblastic leukemia images segmentation on the FPGA’, Adv. Image Video Process., 2016, 3, (3), pp. 8– 17 6Neoh1 S.C., Srisukkham1 W., Zhang L. et al.: ‘An intelligent decision support system for leukaemia diagnosis using microscopic blood images’, Sci. Rep., 2015, 5, p. 14938 7Anto Bennet M., Diana G., Pooja U. et al.: ‘Texture metric driven acute lymphoid leukemia classification using artificial neural networks’, Int. J. Recent Technol. Eng. (IJRTE), 2019, 7, (6S3), pp. 152– 156 8Chatap N., Shibu S.: ‘Analysis of blood samples for counting leukemia cells using support vector machine and nearest neighbour’, J. Comput. Eng., 2014, 16, (5), pp. 79– 87 9Piuri V., Scotti F.: ‘ Morphological classification of blood leucocytes by microscope images’. 2004 IEEE Int. Conf. on Computational Intelligence for Measurement Systems and Applications (2004 CIMSA) , Boston, MA, USA, 2004 10Scotti F.: ‘ Automatic morphological analysis for acute leukemia identification in peripheral blood microscope images’. IEEE Int. Conf. on Computational Intelligence for Measurement Systems and Applications (CIMSA 2005), Giardini Naxos, Italy, 2005 11Scotti F.: ‘ Robust segmentation and measurements techniques of white cells in blood microscope images’. Instrumentation and Measurement Technology Conf., Sorrento, Italy, 2006 12Reta C., Altamirano L., Gonzaleza J.A. et al.: ‘ Segmentation of bone marrow cell images for morphological classification of acute leukemia’. in Proc. of the Twenty-Third Int. Florida Artificial Intelligence Research Society Conf., Daytona Beach, FL, USA, 2010 13Joshi M.D., Karode A.H., Suralkar S.R.: ‘White blood cells segmentation and classification to detect acute leukemia’, Int. J. Emerg. Trends Technol. Comput. Sci., 2013, 2, (3), pp. 147– 151 14Vaghela H.P., Modi H., Pandya M. et al.: ‘Leukemia detection using digital image processing techniques’, Int. J. Appl. Inf. Syst., 2015, 10, pp. 43– 51 15Rejintal A., Aswini N.: ‘Image processing based leukemia cancer cell detection’, Int. J. Eng. Res. Technol. (IJERT), 2016, 4, (22), pp. 471– 474 16Kumar S., Mishra S., Asthana P. et al.: ‘Automated detection of acute leukemia using k-mean clustering algorithm’, in: ‘ Advances in computer and computational sciences’ ( USA, 2018), pp. 655– 670 17Hariprasath S., Dharani T., Santhi M.: ‘Detection of acute lymphocytic leukemia using statistical features’, SSRG Int. J. Electron. Commun. Eng., 2017, pp. 7– 13 18Hegde R., Prasad K., Hebbar H. et al.: ‘Automated decision support system for detection of leukemia from peripheral blood smear images’, J. Digit. Imaging, 2019, 33, pp. 361– 374 19Kumar J.K., Sekhar D.H.: ‘Nucleus and cytoplasm-based segmentation and actor-critic neural network for acute lymphocytic leukaemia detection in single cell blood smear images’, Med. Biol. Eng. Comput., 2020, 58, pp. 171– 186 20Janaki R.: ‘Detection of leukemia in microscopic white blood cell images using Gaussian feature convolutional visual recognition algorithm’, J. Critical Rev., 2020, 7, (3), pp. 173– 180 21Mohammad Z.F., Abdulla A.A.: ‘Thresholding-based white blood cells segmentation from microscopic blood images’, UHD J. Sci. Technol., 2020, 4, pp. 9– 17 22Nikitaeva V., Pronichev A., Polyakov E. et al.: ‘Bone marrow cells recognition methods in the diagnosis of minimal residual disease’, Procedia Comput. Sci., 2020, 169, pp. 353– 358 23Alexandra B., Pavel K., Jan Z.: ‘Automated detection of acute lymphoblastic leukemia from microscopic images based on human visual perception’, J. Front. Bioeng. Biotechnol., 2020, 8, p. 1005 24Cecilia D.R., Loddo A., Puglisi G.: ‘Blood detection and deep learning for leukemic blood image analysis’, Appl. Sci., 2020, 10, p. 1176 25Claro M., Vogado L., Veras R. et al.: ‘ Convolution neural network models for acute leukemia diagnosis’. 2020 Int. Conf. on Systems, Signals and Image Processing (IWSSIP), Bratislava, Slovakia, 2020 26Al-jaboriy S., Sjarif N., Chuprat S. et al.: ‘Acute lymphoblastic leukemia segmentation using local pixel information’, Pattern Recognit. Lett., 2019, 125, pp. 85– 90 27Ziyang Z., Chenrui D., Tao L. et al.: ‘GVFOM: a novel external force for active contour based image segmentation’, Inf. Sci., 2020, 506, pp. 1– 18 28Wang W., Wang Y., Wu Y. et al.: ‘Quantification of full left ventricular metrics via deep regression learning with contour-guidance’, IEEE Access, 2019, 7, pp. 47918– 47928 29Muhammad R., Attique K.M., Majed A. et al.: ‘A sustainable deep learning framework for object recognition using multi-layers deep features fusion and selection’, Sustainability, 2020, 12, p. 5037 30Attique K.M., Ashraf I., Alhaisoni M. et al.: ‘Multimodal brain tumor classification using deep learning and robust feature selection: a machine learning application for radiologists’, Diagnostics, 2020, 10, p. 565 31Zahoor S., Lali I.U., Attique K.M. et al.: ‘Breast cancer detection and classification using traditional computer vision techniques: a comprehensive review’, Curr. Med. Imag., 2020, doi: https://doi.org/10.2174/1573405616666200406110547 32Attique K.M., Kadry S., Alhaisoni M. et al. ‘Computer-aided gastrointestinal diseases analysis from wireless capsule endoscopy: a framework of best features selection’, IEEE Access, 2020, 8, pp. 132850– 132859 33Nazar H.U., Attique K.M., Ullha L.I. et al.: ‘A unified design of ACO and skewness based brain tumor segmentation and classification from MRI scans’, J. Control Eng. Appl. Inf., 2020, 22, pp. 43– 55 34Mishra S., Sharma L., Majhi B. et al.: ‘ Microscopic image classification using DCT for the detection of acute lymphoblastic leukemia (ALL)’. Proc. of Int. Conf. on Computer Vision and Image Processing, Advances in Intelligent Systems and Computing, 2017 Volume14, Issue17December 2020Pages 4435-4440 FiguresReferencesRelatedInformation
Referência(s)