A Comprehensive Overview and Performance Analysisofmri Image Based Brain Tumor Segmentation and Classification Approaches
1Kannika Parameshwari, 2Dr.K.Batri, 3S.Saravanan
1Assistant Professor, Department of ECE, NPR College of Engineering and Technology, Natham, Dindigul.
2Professor, Department of ECE, PSNA College of Engineering and Technology, Dindigul.
3Assistant Professor, Department of ECE, PSNA College of Engineering and Technology, Dindigul.
*Corresponding Author: [email protected]
Abstract.
Brain tumor is a significant brain disease that emerges in view of the accumulation of unvarying cells that becomes or found inner and outer portion of cerebrum. A few imaging methods are utilized to distinguish the tumors in brain region. Among all the imaging strategy, a protected and non-obtrusive imaging method is known as MRI and is primarily used for cerebrum tumor discovery. It is considered as a significant clinical imaging strategy for diagnosing and observing brain tumors, and their capacity for individual treatment. The process of addressing the brain tumor is done from MRI image especially by experienced clinicians. Additionally for more viable treatment and more noteworthy accuracy, clinicians continually endeavor in the computer aided vision based ways to deal with recognize brain tissue from MRI images. In this paper, we represent a relative overview on conventional MRI brain tumor segmentation techniques. Likewise, we enlighten the different calculations utilized in various phases of brain tumor segmentation like pre-processing, extraction and classification.
Keywords: MRI, segmentation, accuracy, preprocessing, feature extraction, classification.
INTRODUCTION:
An objective of brain tumor overview learning is to observe the patient particular information and significant clinical data, and their suggestive attentions. This data implanted inside the several dimensions image information, it may assist and screen the intercessions once the illness has been distinguished and confined, at last prompting information for clinical analysis, organizing, and treatment for sickness. These cycles can be addressed representational as a pyramid. Each and every pyramid steps, significant procedures are needed to measure the information, concentrate, mark, and address the valuable data. Moreover it is important to address the information in higher degree of reflection to acquire significant clinical information or datasets from which clinical determination and conclusion was made. The efficient framework, treatment, representation, and image analysis of the acquired datasets can't be refined without superior performance framework that could be handling by high configuration processors, data storage space, network system, display element of image, just like predefined programs. Improvement and usage of the connected methods require itemized comprehension of the hidden issues, furthermore, comprehension about the obtained information. The feasible extraction of valuable data and individuality contained in various methods of several dimension images may assume a vital part of segmentation of an image.Bhandar et al. (1997) the essential objective of picture division is accustomed to parceling a picture into generally inconsequential regions so much that each territory is spatially adjoining and the inner portion of the pixels are homogeneous regarding a well defined model. This definition represents a huge requirement of most of the division strategies, especially while defining and depicting "peculiar tissue types", consequently the tumors to be isolated are anatomical plans which are as often as possible non-unyielding and complex fit as a shape, wavers colossally in size and position, and show broad changeability from one patient to another. In the specific instance of brain tumors, division contain of separating the distinguishingtissues in tumor such as;tough or vibranttumor, edema, from typical brain tissues, like grey, white matters and CSF. However, brain tumor considers; the sample occurrence of abnormalportionscan be efficientlydetectablelikewise as predicted. By the by,
exact and reproducible division and portrayal of irregularities are not seems to be direct.
Fig No: 1.1 Different modalities of brain MRI.
Fairly as of late different experts in clinical imaging domain and fragile figuring have induced significant changes in f cerebrum tumor division field. The semi-customized and totally modified procedures are the proposed methods and the clinical affirmation of division methodology was depended upon the ease of computation and its degree of assessment. Until better arrangements are proposed, self-programmed or computer aided intelligent techniques will probably be prevailing by and by for quite a while to come, in light of the fact that incorrect transformations are not satisfactory under any conditions. This paper presents an outline of the noteworthyconventional tumor partition strategies in brain area is utilized,once the image has been obtained. The obtainedoutcome of MRI over varioussuggestive imaging methods, this study is focused on tumor segmentation in brain locale. Self-programmed and completely programmed procedures are enlightened.A.R. Deepa et al. (2019), image processing in medical domain, T1- and T2-weighted images are likely to consider as two MRI modalities regularly utilized for representations. These MR cut weightings prefer to the significant sign estimated to make the differentiation in the image. The zones with high fat substance have a short T1 time comparative with water;
T1weighted images might be considered as imagining fat areas. In dissimilarity zones with adequate water level have a short T2 time identified with regions of high fat substance; T2-weighted images are referred as imagining areas of water. A model of FLAIR and weighted of T1, T2 images are appeared in Figure 1. To address the cerebrum tumors, a second T1weighted image is accumulated as often as possible after the infusion of a differentiation agent.
This different set of assessment is exercise to have components whose organization makes a T1 season decrement of close by tissue. In left image: T1-weighted image (light districts addressed areas of fat). A Top right image: T2- weighted (light districts addressed water areas).Prabhjot Kaur Chahal et al. (2020) the beginning module is assortment and securing of images is to be investigated in every stages. Ordinary brain tumor data sets, to be specific BRATS, Harvard clinical assessment, BrainWeb are the ordinarily utilized by the scientists. Besides, these reviews are approved by their methodologies on information stores gathered from different medical and pathology laboratory. The pre-handling module is intended to improve quality of the image visualization by sort out the noise rate. However, the unusual frames make strange/typical tissue separation generally biasing and consequently the exact interpretations. The segmentation is used to enrich the features and supports all the procedure modules, particularly image investigation. Despite the fact that segmentation is significant yet it may not be available once in a while when direct order replace with different destinations. A collection of standard surface highlights alongside other value such as edge and intensity, are ordinarily utilized factors. Powerful learning models are worked by choosing a sub-class of applicable highlights and absolving the unessential value.
Figure 1.2, illustrates the basic operation flow of brain tumor detection.
Fig No: 1.2 Basic block diagram of brain tumor detection
It is effectively accomplished by using standard measurement decrease strategies, for example, guideline part investigation, free segment examination and authoritative connection investigation. At last, the resultant highlights with decreased measurement are utilized for classifier instruction to separate typical brain tissues from strange ones. The execution of the framework is assessed on a few boundaries (like precision, similarity, affectability, dice coefficient, visual correlation). Whenever resultant is recovered, the framework is set to figure and make out the ideal obtained results. To restate, a specialist suggest is constantly liked to have a second assessment for satisfactory conclusion.
Li Liu et al. (2020cerebrum tumors are possibly the most destructive sicknesses with a high death rate. Tumor size and estimation are emotional through the headway cycle. Division is considered as a brain tumor assessment in terms of advancement that limits unmistakable cerebrum tumor improvements, for example, dynamic, edema and some part of tumor degradation tissues from ordinary cerebrum tissue. Mammogram procedure doesn't influences the image processing can be done by some secret tissues in human body and it has a capacity level in terms of tomographic values. Here, the experts might have an expertise in MRI cerebrum to see the internal portion of the brain especially addressing the tumor region. These images are basically seems to be grey and white images which is based on its properties. Also, it is very complex to address the brain tumor image segmentation by MRI image.
However, the existing type of single mode brain MRI tumor addressing method is adaptable and it does not provide enough information about the tumor region. Hence, the Sparse Subspace Clustering technique has been used to examine the exact location of the brain tumor and its accurate dimension Separated and the essential 15 in the Brats 2015 conflict, the precision isn't totally unique, and it is essentially steady somewhere in the range of 10 and 15. To verify the value of noise opposition of the proposed system, this systemincludes6%, 11%, 17%, and 21% of Gaussian noise in terms of examining the various image. Thetestoutcome illustrates, that the proposed systemoutcome has induced noise that is not likely to be affected above the comparable algorithm.
Nelly Gordillo et al. (2019), the process of segmenting the brain tumor is comprises of isolating the various tissues from typical brain tissues: grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF). It contemplates;
the presence of unusual tissues might be effectively distinguishable more often. In any case, precise and reproducible segmentation and portrayal of anomalies are not visibly clear. Previously, frequentclinicians in the domain of medical imaging and promptaddressing have made respectiveoverviews in partition of tumor in brain region. Both self-programmed and completely programmed strategies have been proposed. Clinical recognition of segmentation strategies isbased on the simplicity of the segmentation, and eachposition of image supervision.
Computerized or self-programmedstrategies are most frequently stays inestablished and quite a while,
predominantly in these submission where incorrectperceptive are unacceptable. This paperdescribes the effective and most relevant tumor segmentation strategies in brain region, intended by followingprotection of the image.
Moreover,these reimbursements of MRI image more than other diagnostic technique and this summarization arefocused on tumorpartition in brain region. However, its partially trained and fully trained techniques are emphasized.
Umit Ilhan et al. (2017), the brain disease is a strange cell that occurs in the cerebrum. A clinical MRI is referred as the widelyutilized technique to address and discover the brain tumor. Received pictures by significant medical imaging strategies used toenhancethe picture quality through adapting picture securing methodology. Here, to build up a method for perceptibly perceiving the tissues gets disturbedbecause of its abnormalgrowth. The proposed approach is used to procure a divided tumor locale adequately clear to be seen by the clinical subject matter expert and give them more understanding with respect to the tumor in their decision.. Exactly when taken a gander at, the proposed approach gave an ideal result over the other methodology. The proposed framework created in this examination is a guide for the clinical individuals to analyze the brain malignancy utilizing MRI images. The images received from The Cancer Imaging Archive (TCIA) are utilized in this examination. The proposed system has 94.28% acknowledgment rate on the images with a tumor and 100% acknowledgment rate on the ones without a tumor. The general achievement pace of our proposed system is 96% which shows a superior presentation in examination.
Javeria Amin et al. (2019) the proposed system accomplishes better outcomes on MRI/CT. Cerebrum tumor is maybe the most dazzling sicknesses these days. The gigantic goal of this appraisal work is to introduce another method for the divulgence of tumor. This proposed system precisely sections and social events the warm and bargaining tumor cases. Obvious spatial district methods are applied to overhaul and unequivocally partition the information pictures. Besides, Alex and Google networks are used for demand in which two score vectors are gotten after the softmax layer. Further, both score vectors are combined and given to various classifiers nearby softmax layer. Evaluation of proposed model is done on top clinical picture taking care of and PC helped intervention (MICCAI) challenge datasets i.e., multimodal cerebrum tumor division (BRATS) 2013, 2014, 2015, 2016 and ischemic knock sore division 2018 separately.
Zheshu Jia et al. (2019) the acknowledgment, division, and feature extraction from Magnetic Resonance Imaging pictures of tumor violated locales are seems to be massiveattention; in spite of anydifficult and unrestrainedassignmentprocessed by expert clinicians based on their capacity. Picture taking care of thoughts couldpredict the mixed anatomical collection of the organ of the human parts. A proper confirmation of the cerebrum strangeidentification which is by basically involving some imaging techniques to obtain the result as apositive attempt.However, a novel approach was defined and it is known as Fully Automatic Heterogeneous Segmentation by using Support Vector Machine (FAHS-SVM) was proposed tosegment the tumor in brain region which is purely based on deep learning construction framework. Hence, the unit proportion of thiswholevenous found in cerebral region and it is constructed into MRI imaging with adequate enhancement of another, totally customized computation subject to essential information such as morphological and relaxometry. As far as possible is seen by a basic degree of consistency between life systems and the lining cerebrum tissue. ELM is such a learning tally containing at any rate one layers of covered different focus focuses. Such affiliations are utilized in different areas, including reversal andimage categorization. The MR brainimages, the probabilistic neural network optimization structure are being used to prepare and address the tumor addressing accuracy value in given images.
A mathematical resultant shows practically 98.51% of accuracy.
Guotai Wang et al. (2019) a programmed division of cerebrum tumors region received from expert clinical pictures are referred as huge cumulative value for clinical assessments and choosing treatment based on severity oftumor portion. However, an extending utilization of convolutional neural organization (CNNs) to this errand, however the greater part of them use either 2D associations with for the most part low memory fundamental while sitting above 3D setting, or 3D affiliations manhandling 3D highlights while with gigantic memory usage.
Moreover, existing techniques rarely give weakness information related the division result. We propose a course of CNNs to parcel cerebrum tumors with reformist sub-regions from multi-detached Magnetic Resonance pictures (MRI), and present a 2.5D affiliation that is a tradeoff between memory utilization, model intricacy and
responsive field. In addition, we use test-time amplification to achieve improved division accuracy, which in like manner provides the voxel-wise and expansion weakness in division result sequence. A proper analysis has been done alongwith BraTS 2017 dataset and it is illustrated that our fell design close by 2.5D CNNs was considered as one among the highoperatingtechniquewith respect to BraTS test.In this proposed system, the 98.43% of accuracy rate achieved.
Jianxin Zhang et al. (2020) a programmed segmentation of brain tumors from MRI is a difficult undertaking because of the asymmetrical, unpredictable accomplished great execution. In the mean time, the viability of improving neighborhood reactions for incorporates extraction and recovery has in like manner been showed up in advancing works, which may draw in the better presentation of the cerebrum tumor parcel issue. Proceed by this;
we try to carry the idea fragment into the current U-Net planning to investigate the impacts of adjacent basic reactions on this engineering to investigate the impacts of neighborhood significant reactions on this errand. All the more explicitly, we propose a start to finish 2D brain tumor partition network, i.e., consideration lingering U- Net (AResU-Net), which all the while implants consideration component segment is sufficiently overhaul close by responses of down-reviewing waiting features.The test resultant address, the proposed AResU-Net system overcome its baselines and receivesrelativeimplementationalongside ordinary brain tumor partition strategies and obtained 99.3% of accuracy.
Ramy A. Zeineldin et al. (2020) the DeepSeg structure comprises of a few computerized feature extractors notwithstanding a image extending way. The incorporateassessment resultantwas gained by utilizing2fold cross- endorsement over the 337 planning occurrences of the BraTS-2019 database isolated as pursue: 273 cases for training and 66 for approval. The proposed deep learning designs had the option to precisely perceive tumor regions. In spite of the fact that statistical examination of results is generally close or indistinguishable (like particularity), these outcomes give a significant sign that completely automated deep learning models. A respective Xception encoder illustrated the most excellentvibrantin terms ofsensitivityvalueobtains 0.857, near approximate is 8% more than principal U-Net system, also it may achieve a similar estimation of the unequivocality. This result confirms that point-based methodologies are deficient for evaluating cerebrum tumor division strategy. Thusly, the HD assessments were included to check both the best exactness and execution among all attempted profound encoders. The Mobile-Net addressed the most minimal HD assessment of 10.925 RESULTS AND DISCUSSIONS
Identifying the presence of brain tumors from MRI in a quick, precise, and reproducible way is a difficult issue.
Clinical image handling is an extremely dynamic and quickly developing field that has advanced into a set up discipline. Brain tumor segmentation procedures have effectively shown incredible potential in recognizing and investigating tumors in clinical images and this pattern will without a doubt proceed into what's to come. Clinical image investigation needs to address genuine issues that have been outside the domain of PC vision. These issues come to a great extent from the way that the end frameworks are for the most part utilized by the clinicians. The human factor is fundamental, since any effective arrangement should be acknowledged by a doctor and incorporated into the clinical procedural workflow. This puts solid limitations on the sort of material strategies.
Because of it, there has been a discrepancy between the high level structures introduced in computer aided vision and the low-level techniques utilized by scientists chipping away at genuine clinical application arrangements.
One significant objective in tumor imaging research is to precisely find the disease. Division strategies have been applied by the qualities that permit recognizing tumors from ordinary tissues. At the point when tumors can be recognized from typical tissues by their image force, edge based or locale developing procedures have been utilized. Different tumors are addressed by their dimensions with the goal that a significant-based method was involved for the partition.The performance evaluation of classification algorithms has been calculated with respect to accuracy, sensitivity andspecificity using the following expressions,
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =
𝑇𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 +𝑇𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒𝑇𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 (3.1)
𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 =
𝑇𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑇𝑟𝑢𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣 𝑒+𝐹𝑎𝑙𝑠𝑒 𝑛𝑒𝑎𝑔𝑡𝑖𝑣𝑒(3.2)
𝑆𝑝𝑒𝑐𝑖𝑓𝑖𝑐𝑖𝑡𝑦 =
𝑇𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒𝑇𝑟𝑢𝑒 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 +𝐹𝑎𝑙𝑠𝑒 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 (3.3) Table No: 1.1 Performance metricanalysis of various image segmentation methods
Author Methods Sensitivity
(%)
Specificity (%)
Accuracy (%) Li Liu et al.
(2020)
Sparse Subspace Clustering
97.6 98.2 98.6
Umit Ilhan et al. (2017)
Threshold approach 94.27 99.6 96.3
Javeria Aminet al. (2019)
Score Level Fusion Using Transfer Learning
97.27 96.4 98.1
Zheshu Jiaet al.
(2019)
FAHS-SVM 98.1 96.9 98.51
Guotai Wanget al. (2019)
Cascaded CNN 98.6 99.2 98.43
Jianxin
Zhanget al.
(2020)
AResU-Net 98.9 98.6 99.3
Ramy A.
Zeineldin et al.
(2020)
DeepSeg 98.4 99.8 99.3
Fig No: 1.3 Graphical illustration of Performance metric analysis of various image segmentation methods Table 1.1, illustrates the performance result analysis of various conventional segmentation results. Figure 1.3, represents the graphical illustration of Performance result analysis of various image segmentation methods. Based on the result comparison table, it clearly executes the perfect outcomes of Ramy A. Zeineldin et al. (2020),
97.6 94.27 97.27 98.1 98.6 98.9 98.498.2 99.6 96.4 96.9 99.2 98.6 99.898.6 96.3 98.1 98.51 98.43 99.3 99.3
10 20 30 40 50 60 70 80 90 100
Li Liu et al.
(2020)
Umit Ilhan et al. (2017)
Javeria Amin et al.
(2019)
Zheshu Jia et al. (2019)
Guotai Wang et al.
(2019)
Jianxin Zhang et al.
(2020)
Ramy A.
Zeineldin et al. (2020)
Numerical values (%)
Conventional Methods
Sensitivity (%) Specificity (%) Accuracy (%) Parameter
Deepseg method obtained 98.4% of sensitivity, 99.8% of specificity and 99.3% of accuracy values, when compared with other existing brain tumor segmentation methods, and it provides better results.
CONFLICTS OF INTEREST:
The author have declared no conflicts of interest
REFERENCES
[1] Li Liu, Liang Kuang, and Yunfeng Ji, 2020, “Multimodal MRI Brain Tumor Image Segmentation Using Sparse Subspace Clustering Algorithm”, Computational and Mathematical Methods in Medicine, Volume 2020, Article ID 8620403, 13 pages.
[2] UmitIlhana, AhmetIlhana, 2017, “Brain tumor segmentation based on a new threshold approach”, Procedia Computer Science, Volume 120, 2017, Pages 580-587.
[3] Javeria Amin, Muhammad Sharif, Yasmin Mussarat, Tanzila Saba, 2019, “A New Approach for Brain Tumor Segmentation and Classification Based on Score Level Fusion Using Transfer Learning”, Journal of Medical Systems, 43: 326.
[4] S.Jayachitra, A.Prasanth, 'Multi-Feature Analysis for Automated Brain Stroke Classification Using Weighted Gaussian Naïve Baye’s Classifier', Journal of Circuits, Systems, and Computers, 2021.
[5] Ain, Q, Jaffar, M. A & Choi, T. S. 2014, ‘Fuzzy anisotropic diffusion based segmentation and texture based ensemble classification of brain tumor’. Applied soft computing, vol.21, pp. 330-340.
[6] Aksam Iftikhar, M, Jalil, A, Rathore, S, Ali, A, & Hussain, M. 2013, ‘Brain MRI denoizing and segmentation based on improved adaptive nonlocal means’. International journal of imaging systems and technology, vol.23, no.3, pp. 235-248.
[7] Amato, F, López, A, Peña-Méndez, E. M, Vanhara, P, Hampl, A & Havel, J. 2013, ‘Artificial neural networks in medical diagnosis’.
[8] Bezdek, J. C, Hall, L. O, & Clarke, L. 1993 ‘Review of MR image segmentation techniques using pattern recognition’. Medical physics, vol.20, no.4, pp.10331048.
[9] Bezdek, J. C, Hall, L. O, & Clarke, L. 1993, ‘Review of MR image segmentation techniques using pattern recognition’. Medical physics, vol.20, no.4, pp.10331048.
[10] Bozkurt S, Gimenez F, Burnside ES, Gulkesen KH & Rubin DL. 2016. ‘Using automatically extracted information from mammography reports for decisionsupport’. Journal of biomedical informatics.
Vol.62, pp.224-31.
[11] Cabezas, M, Oliver, A, Lladó, X, Freixenet, J & Cuadra, M. B. 2011 ‘A review of atlas-based segmentation formagnetic resonance brain images’. Computer methods and programs in biomedicine, vol.104, no.3, pp. 158 - 177.
[12] Cai, H, Verma, R, Ou, Y, Lee, S. K., Melhem, E. R & Davatzikos, C. 2007, ‘Probabilistic segmentation of brain tumors based on multi-modality magnetic resonance images’. In Biomedical Imaging: From Nano to Macro, 2007. ISBI 2007. 4th IEEE International Symposium on IEEE, pp. 600-603.
[13] Calabrese, Christopher, Helen Poppleton, Mehmet Kocak, Twala L. Hogg, Christine Fuller, Blair Hamner, Eun Young Oh et al. "A perivascular niche for brain tumor stem cells." Cancer cell 11, no. 1 (2007): 69- 82.
[14] Callaghan, M. F, Freund, P., Draganski, B., Anderson, E., Cappelletti, M., Chowdhury, R & Lutti, A. 2014,
‘Widespread age-related differences in the human brain microstructure revealed by quantitative magnetic resonance imaging’. Neurobiology of aging, vol.35, no.8, pp. 1862-1872.
[15] Chaddad, A, Zinn, P. O & Colen, R. R. 2014, ‘Brain tumor identification using Gaussian Mixture Model features and Decision Trees classifier’. In Information Sciences and Systems (CISS), 2014 48th AnnualConference on IEEE, pp. 1-4.
[16] Charfi, S, Lahmyed, R & Rangarajan, L. 2014, ‘A novel approach for brain tumor detection using neural network’. International Journal of Research in Engineering and Technology, vol.2, pp. 93-104.
[17] Clarke, L. P, Velthuizen, R. P, Camacho, M. A, Heine, J. J, Vaidyanathan, M., Hall, L. O & Silbiger, M. L.
1995 ‘MRI segmentation: methods and applications’. Magnetic resonance imaging, vol.13, no.3, pp.
343-368.
[18] S.Usha Kiruthika, S. Kanaga Suba Raja, V. Balaji, C.J.Raman, S. S. L. Durai Arumugam 2019‘Detection of Tuberculosis in Chest X-rays using U-Net Architecture’, International Journal of Innovative Technology and Exploring Engineering, ISSN: 2278–3075, Volume 9, Issue - 1, pp. 2514-2519
[19] Cuadra, M. B, Pollo, C, Bardera, A, Cuisenaire, O, Villemure, J. G & Thiran, J. P. 2004, ‘Atlas-based segmentation of pathological MR brain images using a model of lesion growth’. IEEE transactions on medical imaging, vol.23, no.10, pp. 1301-1314.
[20] Dahab DA, Ghoniemy SS & Selim GM. 2012. ‘Automated brain tumor detection and identification using image processing and probabilistic neural network techniques’. International journal of image processing and visual communication. Vol.1, no.2, pp.1-8.
[21] A.R.Deepa, W.R. Sam Emmanuel, 2018, “A Comprehensive Review And Analysis On MRI Based Brain Tumor Segmentation”, International Journal Of Scientific & Technology Research, VOLUME 8, ISSUE 10.
[22] Prabhjot Kaur Chahal, Shreelekha Pandey and Shivani Goel, 2020, “Shreelekha Pandey and Shivani Goel A survey on brain tumor detection techniques for MR images”, Multimed Tools Appl 79, pp.
21771–21814.
[23] Sampathkumar, A., S. Murugan, Ahmed A. Elngar, Lalit Garg, R. Kanmani, and A. Christy Jeba Malar. "A Novel Scheme for an IoT-Based Weather Monitoring System Using a Wireless Sensor Network."
In Integration of WSN and IoT for Smart Cities, pp. 181-191. Springer, Cham, 2020.
