|
The Detection of brain abnormality is a complex task. The images captured from the MRI scan machines have numerous information, and it is difficult to segment the appropriate information from the images. Earlier studies have shown various challenges in detecting brain abnormalities through image processing. And one among them is the segmentation of the appropriate region with abnormalities. Thus, the paper proposes a hybrid approach using GAN, K-means clustering, and MobileNet to detect brain abnormalities. Here, the GAN is used to generate fake images from the real images of MR scans. This fake image has been enhanced, and it supports segmenting the abnormal zone from the image using K-means clustering. The segmented region is identified and analyzed for the abnormalities using the MobileNet. Finally, the proposed model could detect abnormalities from the MR images of the brain. The performance metrics for the proposed model are measured and compared, which indicates the improved performance of the proposed model. Introduction Advances in deep learning (Mittal et al., 2019) have transformed medical applications. Through image processing, countless applications are being made. One of the primary applications achieved through image processing (Kailasam & Thiagarajan, 2021) is detecting disease or abnormalities from a part of the Human body (Huang, Shlobin, Lam, & DeCuypere, 2022). Deep learning (Veeramuthu, Meenakshi, & Ashok Kumar, 2019) has opened countless possibilities in detecting diseases earlier and diagnosing the type of disease (Kalaiselvi, Kumar, Subhashri, & Siddharth). These modern techniques have challenges (Nadeem et al., 2020) in detecting complex parts of human body, like our brain. The doctor manually makes the diagnosis in human brain by observing abnormalities in the MR scans (Wadhwa, Bhardwaj, & Verma, 2019) of the brain. The automation of detecting abnormalities in the brain is difficult due to the complex structure of the human brain. Identifying patterns in the human brain for abnormalities depends on various factors (Nazar et al., 2020, Tiwari et al., 2020). The COVID-19 pandemic has raised many questions. Specifically, managing the workforce during an emergency. Many have felt that during such an emergency, automation of predicting disease from the results could improve the problem. This idea made many researchers engage in developing prediction systems for various diseases to diagnose from the CT (Kogilavani et al., 2022, van der Heyden et al., 2019, Woźniak et al., 2021), MRI results (Karayegen and Aksahin, 2021, Tiwari et al., 2020). Different deep learning models are framed for predicting the diseases such as brain tumors (Bhandari, Koppen, & Agzarian, 2020). Yet, there are also numerous challenges in making these prediction models, as their lives are dependent on them. Many recent studies have demonstrated efficient methods for predicting different tumors (Kermi, Andjouh, & Zidane, 2018). Segmentation is a technique where the image is segmented into different small proportions to analyze further for specific information. Considering the complex structure of the human brain, the segmentation technique could assist the model to select the appropriate feature for the classification (Malathi & Sinthia, 2018). And currently, some techniques involve the partial intervention of the doctor (Chaudhari & Kulkarni, 2019) to decide on the diagnosis. The research community preferred MR scan images for the prediction as it has delivered efficient results (Tiwari et al., 2020). The paper is organized as follows, in section II, studies related to the present research area are reviewed. Section III discusses the selection of algorithms for the proposed study, and section IV discusses the methodology for the proposed study. At last, Section V discusses the results of the proposed study, and section VI delivers the conclusion. Section snippets Literature review The research study (Ghassemi, Shoeibi, & Rouhani, 2020) has demonstrated a new deep learning method for tumor identification in Magnetic Resonance (MR) images. The study General Adversarial Network (GAN) has relied on neural networks on generating images from the dataset. The GAN has been used to generate images from the data collected from 233 patients containing 3064 images for Meningioma, Glioma, and Pituitary tumors. The study has used six layer-based neural networks. Initially, the Fully Research gap From the literature review, many studies have discussed the challenge of segmenting (Rehman, Naqvi, Khan, Khan, & Bashir, 2019) the required region from the MR scan images. Also, the features that are selected from the brain comprise numerous information (Malathi & Sinthia, 2018). Selecting appropriate regions from the brain for a quicker segmentation and classification of the abnormality has become a tedious task (Kermi, Andjouh, et al., 2018). In this case, the study specifically mentioned Techniques employed in the study The proposed study includes GAN, K-means clustering, and Mobile-Net V2 architecture. And in this section, the necessity of using the above-mentioned techniques are discussed. Methodology The literature review mentioned that the use of a hybrid approach could improve the model efficiency. So in the current study, GAN, K-means clustering, and MobileNet were combined to make the segmentation efficient. And in this section, the overall flow of the work is discussed. The workflow of the overall methodology is shown in Fig. 2, at first input images, which are considered real images. These real images are fed into the MLAG GAN, which generates an enhanced fake image. Then the enhanced Results and discussion The results obtained from the detection model will be displayed and discussed in this section. Here, the proposed study uses performance metrics to compare with earlier studies to evaluate the model. The images displayed in Fig. 4 shows the obtained results for the abnormality detection for the random five images selected from the dataset. Here, the first image from left to right is the augmented image from the dataset, named MR input image. The Next image is the Generated fake image by the GAN. Conclusion The study aimed to detect the abnormality in the brain from the limited MR images dataset. Hence, the study used GAN to generate a fake image, where the generated fake image has an image enhancement, which is then easily clustered by the K-means clustering. This cluster formation is then segmented as an image which was used as a feature selection to detect the abnormality using Mobile-Net V2 architecture. As a result, the generated fake image exposed better performance in terms of selecting the CRediT authorship contribution statement M. Raja: Conceptualization, Data curation, Formal Analysis, Methodology, Visualization, Writing – original draft. S. Vijayachitra: Conceptualization, Data curation, Methodology, Project Administration, Supervision, validation, Writing – review & editing. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. References (56) A. Wadhwa et al. A review on brain tumor segmentation of MRI images Magnetic Resonance Imaging (2019) A. Tiwari et al. Brain tumor segmentation and classification from magnetic resonance images: Review of selected methods from 2014 to 2019 Pattern Recognition Letters (2020) P. Sriramakrishnan et al. Modified local ternary patterns technique for brain tumour segmentation and volume estimation from MRI multi-sequence scans with GPU CUDA machine Biocybernetics and Biomedical Engineering (2019) M. Soltaninejad et al. Supervised learning based multimodal MRI brain tumour segmentation using texture features from supervoxels Computer Methods and Programs in Biomedicine (2018) L.H. Shehab et al. An efficient brain tumor image segmentation based on deep residual networks (ResNets) Journal of King Saud University-Engineering Sciences (2021) Z.U. Rehman et al. Texture based localization of a brain tumor from MR-images by using a machine learning approach Medical Hypotheses (2020) Z.U. Rehman et al. Fully automated multi-parametric brain tumour segmentation using superpixel based classification Expert systems with applications (2019) H.M. Rai et al. Automatic and accurate abnormality detection from brain MR images using a novel hybrid UnetResNext-50 deep CNN model Biomedical Signal Processing and Control (2021) M. Mittal et al. Deep learning based enhanced tumor segmentation approach for MR brain images Applied Soft Computing (2019) S. Maharjan et al. A novel enhanced softmax loss function for brain tumour detection using deep learning Journal of Neuroscience Methods (2020) G. Karayegen et al. Brain tumor prediction on MR images with semantic segmentation by using deep learning network and 3D imaging of tumor region Biomedical Signal Processing and Control (2021) J. Huang et al. Artificial intelligence applications in pediatric brain tumor imaging: A systematic review World Neurosurgery (2022) M. Hamghalam et al. High tissue contrast image synthesis via multistage attention-GAN: Application to segmenting brain MR scans Neural Networks (2020) N. Ghassemi et al. Deep neural network with generative adversarial networks pre-training for brain tumor classification based on MR images Biomedical Signal Processing and Control (2020) F. Demir et al. A new deep technique using R-CNN model and L1NSR feature selection for brain MRI classification Biomedical Signal Processing and Control (2022) Z.A. Al-Saffar et al. A hybrid approach based on multiple Eigenvalues selection (MES) for the automated grading of a brain tumor using MRI Computer Methods and Programs in Biomedicine (2021) S. Akter et al. COVID-19 detection using deep learning algorithm on chest X-ray images Biology (2021) M. Angulakshmi et al. A review on deep learning architecture and methods for MRI brain tumour segmentation Current Medical Imaging (2021) A. Bhandari et al. Convolutional neural networks for brain tumour segmentation Insights into Imaging (2020) P.K. Chahal et al. Hybrid approaches for brain tumor detection in mr images Paper presented at the international conference on advances in computing and data sciences (2019) P.K. Chahal et al. A survey on brain tumor detection techniques for MR images Multimedia Tools and Applications (2020) A. Chaudhari et al. Semi-automatic unsupervised MR brain tumour segmentation using a simple Bayesian Framework The Imaging Science Journal (2019) J. Chen et al. Adaptive hierarchical dual consistency for semi-supervised left atrium segmentation on cross-domain data IEEE Transactions on Medical Imaging (2021) S. Devunooru et al. Deep learning neural networks for medical image segmentation of brain tumours for diagnosis: A recent review and taxonomy Journal of Ambient Intelligence and Humanized Computing (2021) S. Ghosh et al. Improved U-Net architecture with VGG-16 for brain tumor segmentation Physical and Engineering Sciences in Medicine (2021) P.P. Gumaste et al. A hybrid method for brain tumor detection using advanced textural feature extraction Biomedical and Pharmacology Journal (2020) M.S. Kailasam et al. Detection of lung tumor using dual tree complex wavelet transform and co-active adaptive neuro fuzzy inference system classification approach International Journal of Imaging Systems and Technology (2021) T. Kalaiselvi et al. Classification of leukemia using convolution neural network European Journal of Molecular & Clinical Medicine (2020) Navigate Down View more references Cited by (0) Recommended articles (6) View full text © 2022 Elsevier Ltd. All rights reserved. About ScienceDirect Remote access Shopping cart Advertise Contact and support Terms and conditions Privacy policy We use cookies to help provide and enhance our service and tailor content and ads. By continuing you agree to the use of cookies . Copyright © 2022 Elsevier B.V. or its licensors or contributors. ScienceDirect® is a registered trademark of Elsevier B.V. ScienceDirect® is a registered trademark of Elsevier B.V.
|
