Hybrid Approach for MRI Segmentation using Deep Learning and Machine Learning Algorithms

Hybrid Approach for MRI Segmentation using Deep Learning and Machine Learning Algorithms | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Hybrid Approach for MRI Segmentation using Deep Learning and Machine Learning Algorithms Suresh Kumar Mandala, Neelima Gurrapu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3377680/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Accurate MRI segmentation is a crucial part of modern medical diagnostics and is essential for early disease diagnosis and effective treatment planning. Vision Transformers (ViT), Kernel-Based Convolutional Neural Networks (CNN), and Multi-Class Support Vector Machines (M-SVM) are all presented in this study as part of a novel hybrid approach to MRI segmentation that improves accuracy and efficiency.Our method employs ViT, which rapidly extracts high-level features from MRI patches, in combination with kernel-based convolutional neural networks, which are well-known for their ability to capture intricate patterns in image data. The M-SVM then refines the classification process, separating the pixels into distinct classes that are suggestive of different tissue types, and the segmentation phase begins without any problems. In addition to increasing the accuracy of MRI segmentation, initial findings suggest that this novel method might set an innovative standard for the analysis of medical images. This research has the potential to be an important development in medical imaging, which would significantly advance the current state of the art in healthcare technology by improving the accuracy with which diagnoses are made and the effectiveness of treatment plans. MRI Segmentation Vision Transformers (ViT) Kernel-Based Convolution Neural Networks (CNN) and Multi-Class Support Vector Machines (M-SVM) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION Image segmentation is a process of dividing an image into multiple segments or regions to simplify the image and make it easier to analyze. In medical imaging, image segmentation is used to identify and isolate specific structures or regions of interest, such as tumors, blood vessels, or organs[1] The benefits of using image segmentation and deep learning for brain tumor classification include the ability to automatically extract meaningful features from brain magnetic resonance (MR) images, which offer significantly better performance than traditional machine learning techniques 2. Deep learning-based techniques automatically extract powerful and discriminative deep features from brain MR images, which can improve the accuracy of the classification 2. Additionally, deep learning models can handle high inter and intra shape, texture, and contrast variations, which is a challenging problem for traditional machine learning techniques[2] traditional deep learning methods for multimodal medical image segmentation, such as fully CNNs, suffer from a deficiency of long-range dependencies and bad generalization performance. This means that fully CNNs may not be able to capture all the relevant information in the input images and may not perform well on new, unseen data. combines the strengths of CNNs and Transformers to achieve better performance and generalization in multimodal medical image segmentation. Specifically, the CNNs are used to extract local features from the input images, while the Transformers are used to model long-range dependencies and capture global context. The benefits of this approach include improved accuracy, faster convergence, and better generalization to new data. The authors demonstrate the effectiveness of HybridCTrm on two benchmark datasets and compare it with a fully CNN-based network, showing that HybridCTrm outperforms the fully CNN-based network on most evaluation metrics.[ 3] The Multi-Class Support Vector Machine (M-SVM) is essential for improving MRI segmentation. It essentially divides pixels into different classes, expressing various tissue types as determined by MRI patches, and serves as the final level of refining. This categorization is essential because it enhances the high-level characteristics found by Vision Transformers and the patterns found by Kernel-Based CNNs, allowing for an accurate and thorough segmentation. Additionally, the M-SVM assists in reducing classification mistakes and noise, potentially improving the whole segmentation process' accuracy and dependability and enabling more precise diagnoses and treatment plans. LIETRATURE SURVEY Traditional methods in medical image analysis include feature-based methods, such as texture analysis and shape analysis, and machine learning-based methods, such as support vector machines and deep learning - Deep learning-based methods have shown great success in various medical image analysis tasks, but they require large amounts of labeled data and may not generalize well to new data [4] traditional methods like Dermoscopy are dependent on the expertise of dermatologists, and some computer-aided diagnosis methods may not be accurate or reliable[5]. previous research on multimodal medical image segmentation using deep learning methods. The authors note that traditional deep learning methods, such as fully CNNs, have limitations in capturing long-range dependencies and generalizing to new data. Therefore, recent research has focused on combining CNNs with other architectures, such as Transformers and kernel-based CNNs, to improve performance. Additionally, some studies have explored the use of support vector machines (SVMs) for classification and segmentation tasks. The authors highlight the importance of multimodal imaging, which provides additional information and improves the discriminative power of the network. Overall, the related works section provides a brief overview of the current state of research in multimodal medical image segmentation using deep learning methods[3] Table 1 Brain MRI images Segmentation using various CNN based methods Reference Object Modalities Network Type Data Set Myronenko et al. [6] Brain MRI FCN BRATS2018 Nie et al. [7] Brain MRI 3D FCN Infant brain images Wang et al. [8] Brain MRI FCN ANDI data set and NITRC data set Borne et al. [9] Brain MRI 3D U-Net 62 healthy brain images Casamitjana et al. [10] Brain MRI V-Net BRATS2017 Moeskops et al. [11] Brain MRI GAN MRBrainS13 Rezaei et al. [12] Brain MRI cGAN BRATS 2017 Giacomello et al. [13] Brain MRI SegAN-CAT BRATS2015, BRATS2019 Image Segmentation: Image segmentation is a critical task in medical image analysis, as it involves identifying and separating different regions or structures within an image. In the context of multimodal medical image segmentation, the goal is to assign labels to each pixel of the input images from different modalities. This allows for more accurate and detailed analysis of the images, which can aid in diagnosis, treatment planning, and monitoring of various medical conditions. Deep learning methods, such as CNNs and Transformers, have shown great promise in achieving accurate and efficient image segmentation, particularly in the context of multimodal imaging. By combining these architectures and leveraging the strengths of each, researchers can develop more powerful and effective models for multimodal medical image segmentation. VISION TRANSFORMERS Vision Transformer (ViT) is an upgraded variant of the Transformer model that was originally developed for natural language processing tasks. ViT is a deep learning model that uses a self-attention mechanism to integrate information from different parts of an image. It splits the image into small patches, which are considered sequence tokens, and then flattens them to generate low-dimension embeddings linearly. Finally, the sequence output is passed as an input to the Transformer encoder. ViT is used for image classification tasks, such as identifying skin diseases like Melanoma. It is preferred over traditional Convolutional Neural Networks (CNNs) because it can handle variable-sized information and allows the positional embeddings of the image. ViT also takes less time in training and does not require convolution layers, making it more efficient. CNN and TRANSFERMERS with M-SVM In the the Hybrid approach pr, CNNs and Transformers are used together to address some of the limitations of traditional CNN-based methods for medical image segmentation. CNNs are very good at learning local features from image data, but they may not be as effective at capturing long-range dependencies between different modalities. Transformers, on the other hand, are designed to capture long-range dependencies, but they may not be as effective at learning local features. By combining CNNs and Transformers in a hybrid architecture, the HybridCTrm approach is able to leverage the strengths of both types of networks. The CNNs are used to extract local features from each modality, while the Transformers are used to capture long-range dependencies between different modalities. This allows the network to better integrate information from multiple modalities and produce more accurate segmentation results. First, Vision Transformers and Kernel-Based Convolutional Neural Networks are utilized to identify complicated patterns and details in MRI data, functioning as intelligent eyes capable of detecting detailed details in images. The Multi-Class Support Vector Machine (M-SVM) is then applied. Consider is that classifies these small information into separate categories, such as different tissue kinds or disease signs. It achieves this by finding the correct boundaries which divide these groups in the most logical way.This procedure is similar to drawing the best possible lines on a map to clearly distinguish distinct terrains. This aids in obtaining a more exact interpreting of the MRI images, making sure the difference between various regions is as precise as possible. This not only helps in spotting finer details in images, but also in more consistently diagnosing conditions. The method tries to improve how we evaluate medical images, probably improving the accuracy of diagnoses and treatments. HYBRID Approach for Medical Image Segmentation Data collection The initial step is to import the medical imaging dataset. This is the basic data that can be used for further analysis, processing, and classification can be performed. Preprocessing There are two stages within the preprocessing phase: The first step is collecting all of the required medical images for the dataset. Alternatively, the data can be extended to improve the feature set through a process known as "data augmentation." Feature Extraction Here, the algorithm uses two powerful methods for feature extraction from the image data: Convolution Neural Networks (CNNs) that use a kernel function. In order to recognize patterns and textures at a finer scale, these networks must first extract local features from the image data. Vision Transformers: Vision Transformers process image data in parallel to capture long-range dependencies between various characteristics or modalities contained in the images. Segmentation and Classification The Multi-Class Support Vector Machine (M-SVM) is used once features have been extracted. In part of the classification process, this essential part classifies the features obtained from the previous step into their respective categories. It's possible that these groups stand in for different kinds of tissue or disease indicators. In essence, it ensures accurate and reliable classification by constructing optimal hyperplanes in a high-dimensional feature space. As a result, the precision of the segmentation is greatly enhanced, allowing for more solid diagnoses to be made. Post-processing The algorithm then enters a post-processing stage once the classification and segmentation steps have been completed. As part of this stage, you will be doing:Reconstructing the individual segments into a whole image is called "image reconstruction."Accuracy, sensitivity, specificity, and other metrics are calculated by the algorithm to evaluate segmentation performance. Visualization and Analysis The segmented images and categorization regions are displayed at this stage. We also perform in-depth analyses of the segmented data, which may be useful in medical care and research. Finalization Finally, the segmented image and its related classification metrics are provided as the finished segmentation's output. Algorithm: Read MRI image. . Apply pre-processing techniques: Apply LoG filter for edge enhancement and noise reduction. Apply CLAHE for contrast enhancement. Feature Extraction using Kernel-Based Convolutional Neural Networks: Define window size and matrix size for feature extraction. Choose pixel from the image with distance d, angle θ, and process using defined kernel functions to extract patterns and textures. Feature Integration using Vision Transformers: Utilize transformers to integrate features extracted from CNNs, capturing long-range dependencies and complex patterns. Classification using Multi-Class Support Vector Machine (M-SVM): Utilize M-SVM to classify the integrated features into distinct categories (like different tissue types or disease markers). Evaluate the performance using appropriate metrics and validate the results on different datasets. Results and Discussions This study work experiment is done with the Python language, Medical Segmentation Decathlon (MSD) [36] data set, from which we took Task01_BrainTumour: There are 750 labels in total, and they are split into two groups: Glioma (dead or active tumour) and edoema. It is a regular MRI scan that is done in a hospital. The Performance of the proposed work was calculated by the various measures. They are Sensitivity = TP/TP + FN Specificity = TN/TN + FP Accuracy = TP + TN/TP + TN + FP + FN Table 1 Performance Analysis of Existing Work and Proposed Work Measures K-Means SVM CNN Proposed Work Sensitivity 84.6 83.1 89.4 95.3 Specificity 86.4 84.3 91.7 94.6 Accuracy 85.3 83.6 90.6 94.8 Error Rate: The number of instances that a decision model has erroneously labelled a pattern. Error rate of proposed method compared to prior work is shown in Figure . Conclusion In this study, we developed a method for improved medical picture analysis by combining Vision Transformers, Kernel-Based Convolutional Neural Networks, and Multi-Class Support Vector Machines (M-SVM). We used LoG and CLAHE for pre-processing to improve image quality, allowing for more precise segmentation and classification. Collectively, proven to be an effective tool for analysing complex MRI patterns, which in turn prepares the way for quicker, more precise diagnoses. This novel method has great promise for the future of medical image processing, as it may lead to improved diagnosis accuracy. References Saeed Iqbal1,2 · Adnan N. Qureshi1 · Jianqiang Li2,3 · Tariq Mahmood. (2023). On the Analyses of Medical Images Using Traditional Machine Learning Techniques and Convolutional Neural Networks. Vol.:(0123456789)1 3Archives of Computational Methods in Engineering . -(30), p.p3173–3233. Jaeyong Kang 1 , Zahid Ullah 1 and Jeonghwan Gwak 1,2,3,4,*. (2021). MRI-Based Brain Tumor Classification Using Ensemble of Deep Features and Machine Learning Classifiers. MDPI sensors . -(-), pp.p2-21. Qixuan Sun , 1 , 2 Nianhua Fang, 1 , 2 Zhuo Liu , 3 Liang Zhao , 1 , 2 Youpeng W. (2021). HybridCTrm: Bridging CNN and Transformer for Multimodal Brain Image Segmentation. Journal of Healthcare Engineering . 2021(-), pp.p1-10 Kelei He 1,2,#, Chen Gan 2,#, Zhuoyuan Li 1,2,#, Islem Rekik 3,4,#, Zihao Yin 2,. (2022). Transformers in medical image analysis. Intelligent Medicine . -(-), pp.p59-78. Vikas Kumar Roy ,asu Thakur,et al. (2023). Vision Transformer Framework Approach For Melanoma Skin Disease Identication. Research Squeare . -(-), pp.p1-12. Myronenko, A. 3D MRI brain tumor segmentation using autoencoder regularization. In Proceedings of the International MICCAI Brainlesion Workshop, Shenzhen, China, 17 October 2018; pp. 311–320. Nie, D.; Wang, L.; Adeli, E.; Lao, C.; Lin, W.; Shen, D. 3-D fully convolutional networks for multimodal isointense infant brain image segmentation. IEEE Trans. Cybern. 2019, 49, 1123–1136. [CrossRef] [PubMed] Wang, S.; Yi, L.; Chen, Q.; Meng, Z.; Dong, H.; He, Z. Edge-aware Fully Convolutional Network with CRF-RNN Layer for Hippocampus Segmentation. In Proceedings of the 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference (ITAIC), Chongqing, China, 24–26 May 2019; pp. 803–806. Borne, L.; Rivière, D.; Mangin, J.F. Combining 3D U-Net and bottom-up geometric constraints for automatic cortical sulci recognition. In Proceedings of the International Conference on Medical Imaging with Deep Learning, London, UK, 8–10 July 2019. Casamitjana, A.; Catà, M.; Sánchez, I.; Combalia, M.; Vilaplana, V. Cascaded V-Net using ROI masks for brain tumor segmentation. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 381–391. Moeskops, P.; Veta, M.; Lafarge, M.W.; Eppenhof, K.A.J.; Pluim, J.P.W. Adversarial training and dilated convolutions for brain MRI segmentation. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support; Springer: Cham, Switzerland, 2017; pp. 56–64. Rezaei, M.; Harmuth, K.; Gierke, W.; Kellermeier, T.; Fischer, M.; Yang, H.; Meinel, C. A conditional adversarial network for semantic segmentation of brain tumor. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 241–252. Giacomello, E.; LoIacono, D.; Mainardi, L. Brain MRI Tumor Segmentation with Adversarial Networks. arXiv 2019, arXiv:1910.02717. Hesamian, M.H.; Jia, W.; He, X.; Kennedy, P. Deep learning techniques for medical image segmentation: Achievements and challenges. J. Digit. Imaging 2019, 32, 582–596. [CrossRef] [PubMed] Altaf, F.; Islam, S.M.S.; Akhtar, N.; Nanjua, N.K. Going deep in medical image analysis: Concepts, methods, challenges, and future directions. IEEE Access 2019, 7, 99540–99572. [CrossRef] Hu, P.; Cao, Y.; Wang, W.; Wei, B. Computer Assisted Three-Dimensional Reconstruction for Laparoscopic Resection in Adult Teratoma. J. Med. Imaging Health Inform. 2019, 9, 956–961. [CrossRef] Ess, A.; Müller, T.; Grabner, H.; Van Gool, L. Segmentation-Based Urban Traffic Scene Understanding. BMVC 2009, 1, 2. Geiger, A.; Lenz, P.; Urtasun, R. Are we ready for autonomous driving? The kitti vision benchmark suite. In Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, RI, USA, 16–12 June 2012; pp. 3354–3361. Ma, Z.; Tavares, J.M.R.S.; Jorge, R.M.N. A review on the current segmentation algorithms for medical images. In Proceedings of the 1st International Conference on Imaging Theory and Applications, Lisbon, Portugal, 5–8 February 2009. Ferreira, A.; Gentil, F.; Tavares, J.M.R.S. Segmentation algorithms for ear image data towards biomechanical studies. Comput. Methods Biomech. Biomed. Eng. 2014, 17, 888–904. [CrossRef] Ma, Z.; Tavares, J.M.R.S.; Jorge, R.N.; Mascarenhas, T. A review of algorithms for medical image segmentation and their applications to the female pelvic cavity. Comput. Methods Biomech. Biomed. Eng. 2010, 13, 235–246. [CrossRef] Xu, A.; Wang, L.; Feng, S.; Qu, Y. Threshold-based level set method of image segmentation. In Proceedings of the Third International Conference on Intelligent Networks and Intelligent Systems, Shenyang, China, 1–3 November 2010; pp. 703–706. Cigla, C.; Alatan, A.A. Region-based image segmentation via graph cuts. In Proceedings of the 2008 15th IEEE International Conference on Image Processing, San Diego, CA, USA, 12–15 October 2008; pp. 2272–2275. Yu-Qian, Z.; Wei-Hua, G.; Zhen-Cheng, C.; Tang, J.-T.; Li, L.-Y. Medical images edge detection based on mathematical morphology. In Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, Shanghai, China, 17–18 January 2006; pp. 6492–6495. He, K.; Gkioxari, G.; Dollár, P.; Girschik, R. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 22–29 October 2017; pp. 2961–2969. Lin, G.; Milan, A.; Shen, C.; Reid, I. Refinenet: Multi-path refinement networks for high-resolution semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 1925–1934. Noh, H.; Hong, S.; Han, B. Learning deconvolution network for semantic segmentation. In Proceedings of the IEEE International Conference on Computer Vision, Las Condes, Chile, 11–18 December 2015; pp. 1520–1528. Gu, J.; Wang, Z.; Kuen, J.; Ma, L.; Shshroudy, A.; Shuai, B.; Liu, I.; Wang, X.; Wang, G.; Cai, J.; et al. Recent advances in convolutional neural networks. Pattern Recognit. 2018, 77, 354–377. [CrossRef] Hubel, D.H.; Wiesel, T.N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. J. Physiol. 1962, 160, 106. [CrossRef] [PubMed] Fukushima, K.; Miyake, S. Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition. In Competition and Cooperation in Neural Nets; Springer: Berlin, Germany, 1982; pp. 267–285. Lécun, Y.; Bottou, L.; Bengio, Y.; Haffner, P. Gradient-based learning applied to document recognition. IEEE 1998, 86, 2278–2324. [CrossRef] Krizhevsky, A.; Sutskever, I.; Hinton, G.E. Imagenet classification with deep convolutional neural networks. Adv. Neural Inf. Process. Syst. 2012, 60, 1097–1105. [CrossRef] He, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 16 June–1 July 2016; pp. 770–778. Simonyan, K.; Zisserman, A. Very deep convolutional networks for large-scale image recognition. arXiv 2014, arXiv:1409.1556. Qiu, Z.; Yao, T.; Mei, T. Learning spatio-temporal representation with pseudo-3d residual networks. In Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 22–29 October 2017; pp. 5533–5541. Long, J.; Shelhamer, E.; Darrell, T. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, 7–12 June 2015; pp. 3431–3440. Myronenko, A. 3D MRI brain tumor segmentation using autoencoder regularization. In Proceedings of the International MICCAI Brainlesion Workshop, Shenzhen, China, 17 October 2018; pp. 311–320. Nie, D.; Wang, L.; Adeli, E.; Lao, C.; Lin, W.; Shen, D. 3-D fully convolutional networks for multimodal isointense infant brain image segmentation. IEEE Trans. Cybern. 2019, 49, 1123–1136. [CrossRef] [PubMed] Wang, S.; Yi, L.; Chen, Q.; Meng, Z.; Dong, H.; He, Z. Edge-aware Fully Convolutional Network with CRF-RNN Layer for Hippocampus Segmentation. In Proceedings of the 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference (ITAIC), Chongqing, China, 24–26 May 2019; pp. 803–806. Borne, L.; Rivière, D.; Mangin, J.F. Combining 3D U-Net and bottom-up geometric constraints for automatic cortical sulci recognition. In Proceedings of the International Conference on Medical Imaging with Deep Learning, London, UK, 8–10 July 2019. Casamitjana, A.; Catà, M.; Sánchez, I.; Combalia, M.; Vilaplana, V. Cascaded V-Net using ROI masks for brain tumor segmentation. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 381–391. Moeskops, P.; Veta, M.; Lafarge, M.W.; Eppenhof, K.A.J.; Pluim, J.P.W. Adversarial training and dilated convolutions for brain MRI segmentation. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support; Springer: Cham, Switzerland, 2017; pp. 56–64. Rezaei, M.; Harmuth, K.; Gierke, W.; Kellermeier, T.; Fischer, M.; Yang, H.; Meinel, C. A conditional adversarial network for semantic segmentation of brain tumor. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 241–252. Giacomello, E.; LoIacono, D.; Mainardi, L. Brain MRI Tumor Segmentation with Adversarial Networks. arXiv 2019, arXiv:1910.02717. Simpson, A.L.; Antonelli, M.; Bakas, S.; Bilello, M.; Farahani, K.; Van Ginneken, B.; Kopp-Schneider, A.; Landman, B.A.; Litjens, G.; Menze, B.; et al. A large annotated medical image dataset for the development and evaluation of segmentation algorithms. arXiv 2019, arXiv:1902.09063 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3377680","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":236115609,"identity":"01344fc3-5546-4be4-8ea8-395d30edc638","order_by":0,"name":"Suresh Kumar Mandala","email":"","orcid":"","institution":"SR University","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Suresh","middleName":"Kumar","lastName":"Mandala","suffix":""},{"id":236115610,"identity":"02405f86-b932-47a6-8a10-bd6cd6c8d727","order_by":1,"name":"Neelima Gurrapu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYLCCBwYMDPz8zQeATAkZ4rQkALVIzjiWANLCQ6QWIDY4kGMAYhPWIh/dfPBBQsE2ecmGM59f3aix4GFgP3x0Az4thneOJRskGNw27Gfu3WadcwzoMJ60tBt4tczIMZMAamGc2XB2m3EOG1CLBI8ZAS35338AtdhvOJDzzDjnHxFa5CVy2IAhdjsRqIX5cW4bEVoMJNKMQQ5LnjnjmBlzbp8EDxshv8jPSH744cOf27b9/M2PP+d8q5PjZz98DL8tBxBsNgkwiU852JYGBJv5AyHVo2AUjIJRMDIBAODWTPbHnChwAAAAAElFTkSuQmCC","orcid":"","institution":"SR University","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Neelima","middleName":"","lastName":"Gurrapu","suffix":""}],"badges":[],"createdAt":"2023-09-22 17:29:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3377680/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3377680/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":44044686,"identity":"3a120be9-8d82-4f32-aada-ac17cd44f39d","added_by":"auto","created_at":"2023-10-03 22:41:34","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":88797,"visible":true,"origin":"","legend":"\u003cp\u003eMRI Brain Tumor image Slices in different dimensions\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-3377680/v1/39e47c930142b6094747556a.png"},{"id":44044688,"identity":"999e362d-5475-4c6c-a37a-1c8e008c7a4c","added_by":"auto","created_at":"2023-10-03 22:41:34","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":135113,"visible":true,"origin":"","legend":"\u003cp\u003eMRI Brain Tumor image Structure Examples\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-3377680/v1/737cf19fe34ccf0e0b6c16c6.png"},{"id":44044689,"identity":"04959a69-8e9d-4def-838d-b1bc0afffef9","added_by":"auto","created_at":"2023-10-03 22:41:34","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":133605,"visible":true,"origin":"","legend":"\u003cp\u003eFigure: MRI Brain Tumor image various Appearances\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-3377680/v1/cba6fc0ad18a9540b67509f6.png"},{"id":44044690,"identity":"77228d58-5ca7-4a78-9679-6f36069e14d3","added_by":"auto","created_at":"2023-10-03 22:41:34","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":203453,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 1: architecture:\u003c/p\u003e","description":"","filename":"12.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3377680/v1/eb0bbd8607a3a52708190693.jpg"},{"id":44044687,"identity":"f69c63c5-4329-4198-a150-60f3fb020fe6","added_by":"auto","created_at":"2023-10-03 22:41:34","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":22025,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 8: Error Rate of Proposed work versus Existing Work\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-3377680/v1/339aecdb7223e42e93423ec9.png"},{"id":48936966,"identity":"eb880a87-7874-4426-91f3-7e5322957af0","added_by":"auto","created_at":"2023-12-29 00:52:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":647421,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3377680/v1/d3e5adfe-59cc-42b2-9c63-bcd4d9d1b9f4.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Hybrid Approach for MRI Segmentation using Deep Learning and Machine Learning Algorithms","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eImage segmentation is a process of dividing an image into multiple segments or regions to simplify the image and make it easier to analyze. In medical imaging, image segmentation is used to identify and isolate specific structures or regions of interest, such as tumors, blood vessels, or organs[1] The benefits of using image segmentation and deep learning for brain tumor classification include the ability to automatically extract meaningful features from brain magnetic resonance (MR) images, which offer significantly better performance than traditional machine learning techniques 2. Deep learning-based techniques automatically extract powerful and discriminative deep features from brain MR images, which can improve the accuracy of the classification 2. Additionally, deep learning models can handle high inter and intra shape, texture, and contrast variations, which is a challenging problem for traditional machine learning techniques[2]\u0026nbsp;traditional deep learning methods for multimodal medical image segmentation, such as fully CNNs, suffer from a deficiency of long-range dependencies and bad generalization performance. This means that fully CNNs may not be able to capture all the relevant information in the input images and may not perform well on new, unseen data. combines the strengths of CNNs and Transformers to achieve better performance and generalization in multimodal medical image segmentation. Specifically, the CNNs are used to extract local features from the input images, while the Transformers are used to model long-range dependencies and capture global context. The benefits of this approach include improved accuracy, faster convergence, and better generalization to new data. The authors demonstrate the effectiveness of HybridCTrm on two benchmark datasets and compare it with a fully CNN-based network, showing that HybridCTrm outperforms the fully CNN-based network on most evaluation metrics.[ 3]\u003c/p\u003e\n\u003cp\u003eThe Multi-Class Support Vector Machine (M-SVM) is essential for improving MRI segmentation. It essentially divides pixels into different classes, expressing various tissue types as determined by MRI patches, and serves as the final level of refining. This categorization is essential because it enhances the high-level characteristics found by Vision Transformers and the patterns found by Kernel-Based CNNs, allowing for an accurate and thorough segmentation. Additionally, the M-SVM assists in reducing classification mistakes and noise, potentially improving the whole segmentation process' accuracy and dependability and enabling more precise diagnoses and treatment plans.\u003c/p\u003e"},{"header":"LIETRATURE SURVEY","content":"\u003cp\u003eTraditional methods in medical image analysis include feature-based methods, such as texture analysis and shape analysis, and machine learning-based methods, such as support vector machines and deep learning - Deep learning-based methods have shown great success in various medical image analysis tasks, but they require large amounts of labeled data and may not generalize well to new data [4]\u0026nbsp;traditional methods like Dermoscopy are dependent on the expertise of dermatologists, and some computer-aided diagnosis methods may not be accurate or reliable[5].\u003c/p\u003e\n\u003cp\u003eprevious research on multimodal medical image segmentation using deep learning methods. The authors note that traditional deep learning methods, such as fully CNNs, have limitations in capturing long-range dependencies and generalizing to new data. Therefore, recent research has focused on combining CNNs with other architectures, such as Transformers and kernel-based CNNs, to improve performance. Additionally, some studies have explored the use of support vector machines (SVMs) for classification and segmentation tasks. The authors highlight the importance of multimodal imaging, which provides additional information and improves the discriminative power of the network. Overall, the related works section provides a brief overview of the current state of research in multimodal medical image segmentation using deep learning methods[3]\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eBrain MRI images Segmentation using various CNN based methods\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eReference\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eObject\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModalities\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNetwork Type\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eData Set\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMyronenko et al. [6]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFCN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBRATS2018\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNie et al. [7]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3D FCN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eInfant brain images\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWang et al. [8]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFCN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eANDI data set and NITRC data set\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBorne et al. [9]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3D U-Net\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e62 healthy brain images\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCasamitjana et al. [10]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eV-Net\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBRATS2017\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMoeskops et al. [11]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGAN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRBrainS13\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRezaei et al. [12]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ecGAN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBRATS 2017\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGiacomello et al. [13]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBrain\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMRI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSegAN-CAT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBRATS2015, BRATS2019\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eImage Segmentation:\u003c/h2\u003e\n \u003cp\u003eImage segmentation is a critical task in medical image analysis, as it involves identifying and separating different regions or structures within an image. In the context of multimodal medical image segmentation, the goal is to assign labels to each pixel of the input images from different modalities. This allows for more accurate and detailed analysis of the images, which can aid in diagnosis, treatment planning, and monitoring of various medical conditions. Deep learning methods, such as CNNs and Transformers, have shown great promise in achieving accurate and efficient image segmentation, particularly in the context of multimodal imaging. By combining these architectures and leveraging the strengths of each, researchers can develop more powerful and effective models for multimodal medical image segmentation.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"VISION TRANSFORMERS","content":"\u003cp\u003eVision Transformer (ViT) is an upgraded variant of the Transformer model that was originally developed for natural language processing tasks. ViT is a deep learning model that uses a self-attention mechanism to integrate information from different parts of an image. It splits the image into small patches, which are considered sequence tokens, and then flattens them to generate low-dimension embeddings linearly. Finally, the sequence output is passed as an input to the Transformer encoder. ViT is used for image classification tasks, such as identifying skin diseases like Melanoma. It is preferred over traditional Convolutional Neural Networks (CNNs) because it can handle variable-sized information and allows the positional embeddings of the image. ViT also takes less time in training and does not require convolution layers, making it more efficient.\u003c/p\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003eCNN and TRANSFERMERS with M-SVM\u003c/h2\u003e\n \u003cp\u003eIn the the Hybrid approach pr, CNNs and Transformers are used together to address some of the limitations of traditional CNN-based methods for medical image segmentation. CNNs are very good at learning local features from image data, but they may not be as effective at capturing long-range dependencies between different modalities. Transformers, on the other hand, are designed to capture long-range dependencies, but they may not be as effective at learning local features. By combining CNNs and Transformers in a hybrid architecture, the HybridCTrm approach is able to leverage the strengths of both types of networks. The CNNs are used to extract local features from each modality, while the Transformers are used to capture long-range dependencies between different modalities. This allows the network to better integrate information from multiple modalities and produce more accurate segmentation results.\u003c/p\u003e\n \u003cp\u003eFirst, Vision Transformers and Kernel-Based Convolutional Neural Networks are utilized to identify complicated patterns and details in MRI data, functioning as intelligent eyes capable of detecting detailed details in images.\u003c/p\u003e\n \u003cp\u003eThe Multi-Class Support Vector Machine (M-SVM) is then applied. Consider is that classifies these small information into separate categories, such as different tissue kinds or disease signs. It achieves this by finding the correct boundaries which divide these groups in the most logical way.This procedure is similar to drawing the best possible lines on a map to clearly distinguish distinct terrains. This aids in obtaining a more exact interpreting of the MRI images, making sure the difference between various regions is as precise as possible. This not only helps in spotting finer details in images, but also in more consistently diagnosing conditions. The method tries to improve how we evaluate medical images, probably improving the accuracy of diagnoses and treatments.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003eHYBRID Approach for Medical Image Segmentation\u003c/h2\u003e\n \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\n \u003ch2\u003eData collection\u003c/h2\u003e\n \u003cp\u003eThe initial step is to import the medical imaging dataset. This is the basic data that can be used for further analysis, processing, and classification can be performed.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003ePreprocessing\u003c/h2\u003e\n \u003cp\u003eThere are two stages within the preprocessing phase:\u003c/p\u003e\n \u003cp\u003eThe first step is collecting all of the required medical images for the dataset.\u003c/p\u003e\n \u003cp\u003eAlternatively, the data can be extended to improve the feature set through a process known as \u0026quot;data augmentation.\u0026quot;\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003eFeature Extraction\u003c/h2\u003e\n \u003cp\u003eHere, the algorithm uses two powerful methods for feature extraction from the image data:\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eConvolution Neural Networks (CNNs)\u003c/strong\u003e that use a kernel function. In order to recognize patterns and textures at a finer scale, these networks must first extract local features from the image data.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eVision Transformers:\u0026nbsp;\u003c/strong\u003eVision Transformers process image data in parallel to capture long-range dependencies between various characteristics or modalities contained in the images.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003eSegmentation and Classification\u003c/h2\u003e\n \u003cp\u003eThe Multi-Class Support Vector Machine (M-SVM) is used once features have been extracted. In part of the classification process, this essential part classifies the features obtained from the previous step into their respective categories. It\u0026apos;s possible that these groups stand in for different kinds of tissue or disease indicators. In essence, it ensures accurate and reliable classification by constructing optimal hyperplanes in a high-dimensional feature space. As a result, the precision of the segmentation is greatly enhanced, allowing for more solid diagnoses to be made.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003ePost-processing\u003c/h2\u003e\n \u003cp\u003eThe algorithm then enters a post-processing stage once the classification and segmentation steps have been completed. As part of this stage, you will be doing:Reconstructing the individual segments into a whole image is called \u0026quot;image reconstruction.\u0026quot;Accuracy, sensitivity, specificity, and other metrics are calculated by the algorithm to evaluate segmentation performance.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eVisualization and Analysis\u003c/h2\u003e\n \u003cp\u003eThe segmented images and categorization regions are displayed at this stage. We also perform in-depth analyses of the segmented data, which may be useful in medical care and research.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eFinalization\u003c/h2\u003e\n \u003cp\u003eFinally, the segmented image and its related classification metrics are provided as the finished segmentation\u0026apos;s output.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eAlgorithm:\u003c/h2\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eRead MRI image.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003e. Apply pre-processing techniques:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eApply LoG filter for edge enhancement and noise reduction.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eApply CLAHE for contrast enhancement.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eFeature Extraction using Kernel-Based Convolutional Neural Networks:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eDefine window size and matrix size for feature extraction.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eChoose pixel from the image with distance d, angle \u0026theta;, and process using defined kernel functions to extract patterns and textures.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eFeature Integration using Vision Transformers:\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eUtilize transformers to integrate features extracted from CNNs, capturing long-range dependencies and complex patterns.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eClassification using Multi-Class Support Vector Machine (M-SVM):\u003c/p\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eUtilize M-SVM to classify the integrated features into distinct categories (like different tissue types or disease markers).\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eEvaluate the performance using appropriate metrics and validate the results on different datasets.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e"},{"header":"Results and Discussions","content":"\u003cp\u003eThis study work experiment is done with the Python language, Medical Segmentation Decathlon (MSD) [36] data set, from which we took Task01_BrainTumour: There are 750 labels in total, and they are split into two groups: Glioma (dead or active tumour) and edoema. It is a regular MRI scan that is done in a hospital.\u003c/p\u003e\n\u003cp\u003eThe Performance of the proposed work was calculated by the various measures. They are\u003c/p\u003e\n\u003cp\u003eSensitivity\u0026thinsp;=\u0026thinsp;TP/TP\u0026thinsp;+\u0026thinsp;FN\u003c/p\u003e\n\u003cp\u003eSpecificity\u0026thinsp;=\u0026thinsp;TN/TN\u0026thinsp;+\u0026thinsp;FP\u003c/p\u003e\n\u003cp\u003eAccuracy\u0026thinsp;=\u0026thinsp;TP\u0026thinsp;+\u0026thinsp;TN/TP\u0026thinsp;+\u0026thinsp;TN\u0026thinsp;+\u0026thinsp;FP\u0026thinsp;+\u0026thinsp;FN\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003ctable id=\"Tab2\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003ePerformance Analysis of Existing Work and Proposed Work\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eMeasures\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eK-Means\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSVM\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCNN\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eProposed Work\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSensitivity\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e84.6\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e83.1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e89.4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e95.3\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSpecificity\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e86.4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e84.3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e91.7\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e94.6\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAccuracy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e85.3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e83.6\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e90.6\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e94.8\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eError Rate: The number of instances that a decision model has erroneously labelled a pattern. Error rate of proposed method compared to prior work is shown in Figure .\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this study, we developed a method for improved medical picture analysis by combining Vision Transformers, Kernel-Based Convolutional Neural Networks, and Multi-Class Support Vector Machines (M-SVM). We used LoG and CLAHE for pre-processing to improve image quality, allowing for more precise segmentation and classification. Collectively, proven to be an effective tool for analysing complex MRI patterns, which in turn prepares the way for quicker, more precise diagnoses. This novel method has great promise for the future of medical image processing, as it may lead to improved diagnosis accuracy.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSaeed Iqbal1,2 \u0026middot; Adnan N. Qureshi1 \u0026middot; Jianqiang Li2,3 \u0026middot; Tariq Mahmood. (2023). On the Analyses of Medical Images Using Traditional Machine Learning Techniques and Convolutional Neural Networks. \u003cem\u003eVol.:(0123456789)1 3Archives of Computational Methods in Engineering\u003c/em\u003e. -(30), p.p3173\u0026ndash;3233.\u003c/li\u003e\n\u003cli\u003eJaeyong Kang 1 , Zahid Ullah 1 and Jeonghwan Gwak 1,2,3,4,*. (2021). MRI-Based Brain Tumor Classification Using Ensemble of Deep Features and Machine Learning Classifiers. \u003cem\u003eMDPI sensors\u003c/em\u003e. -(-), pp.p2-21.\u003c/li\u003e\n\u003cli\u003eQixuan Sun , 1 , 2 Nianhua Fang, 1 , 2 Zhuo Liu , 3 Liang Zhao , 1 , 2 Youpeng W. (2021). HybridCTrm: Bridging CNN and Transformer for Multimodal Brain Image Segmentation. \u003cem\u003eJournal of Healthcare Engineering\u003c/em\u003e. 2021(-), pp.p1-10\u003c/li\u003e\n\u003cli\u003eKelei He 1,2,#, Chen Gan 2,#, Zhuoyuan Li 1,2,#, Islem Rekik 3,4,#, Zihao Yin 2,. (2022). Transformers in medical image analysis. \u003cem\u003eIntelligent Medicine\u003c/em\u003e. -(-), pp.p59-78.\u003c/li\u003e\n\u003cli\u003eVikas Kumar Roy ,asu Thakur,et al. (2023). Vision Transformer Framework Approach For Melanoma Skin Disease Identication. \u003cem\u003eResearch Squeare\u003c/em\u003e. -(-), pp.p1-12.\u003c/li\u003e\n\u003cli\u003eMyronenko, A. 3D MRI brain tumor segmentation using autoencoder regularization. In Proceedings of the International MICCAI Brainlesion Workshop, Shenzhen, China, 17 October 2018; pp. 311\u0026ndash;320.\u003c/li\u003e\n\u003cli\u003eNie, D.; Wang, L.; Adeli, E.; Lao, C.; Lin, W.; Shen, D. 3-D fully convolutional networks for multimodal isointense infant brain image segmentation. IEEE Trans. Cybern. 2019, 49, 1123\u0026ndash;1136. [CrossRef] [PubMed] \u003c/li\u003e\n\u003cli\u003eWang, S.; Yi, L.; Chen, Q.; Meng, Z.; Dong, H.; He, Z. Edge-aware Fully Convolutional Network with CRF-RNN Layer for Hippocampus Segmentation. In Proceedings of the 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference (ITAIC), Chongqing, China, 24\u0026ndash;26 May 2019; pp. 803\u0026ndash;806. \u003c/li\u003e\n\u003cli\u003eBorne, L.; Rivi\u0026egrave;re, D.; Mangin, J.F. Combining 3D U-Net and bottom-up geometric constraints for automatic cortical sulci recognition. In Proceedings of the International Conference on Medical Imaging with Deep Learning, London, UK, 8\u0026ndash;10 July 2019.\u003c/li\u003e\n\u003cli\u003eCasamitjana, A.; Cat\u0026agrave;, M.; S\u0026aacute;nchez, I.; Combalia, M.; Vilaplana, V. Cascaded V-Net using ROI masks for brain tumor segmentation. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 381\u0026ndash;391. \u003c/li\u003e\n\u003cli\u003eMoeskops, P.; Veta, M.; Lafarge, M.W.; Eppenhof, K.A.J.; Pluim, J.P.W. Adversarial training and dilated convolutions for brain MRI segmentation. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support; Springer: Cham, Switzerland, 2017; pp. 56\u0026ndash;64.\u003c/li\u003e\n\u003cli\u003eRezaei, M.; Harmuth, K.; Gierke, W.; Kellermeier, T.; Fischer, M.; Yang, H.; Meinel, C. A conditional adversarial network for semantic segmentation of brain tumor. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 241\u0026ndash;252. \u003c/li\u003e\n\u003cli\u003eGiacomello, E.; LoIacono, D.; Mainardi, L. Brain MRI Tumor Segmentation with Adversarial Networks. arXiv 2019, arXiv:1910.02717.\u003c/li\u003e\n\u003cli\u003eHesamian, M.H.; Jia, W.; He, X.; Kennedy, P. Deep learning techniques for medical image segmentation: Achievements and challenges. J. Digit. Imaging 2019, 32, 582\u0026ndash;596. [CrossRef] [PubMed] \u003c/li\u003e\n\u003cli\u003eAltaf, F.; Islam, S.M.S.; Akhtar, N.; Nanjua, N.K. Going deep in medical image analysis: Concepts, methods, challenges, and future directions. IEEE Access 2019, 7, 99540\u0026ndash;99572. [CrossRef]\u003c/li\u003e\n\u003cli\u003eHu, P.; Cao, Y.; Wang, W.; Wei, B. Computer Assisted Three-Dimensional Reconstruction for Laparoscopic Resection in Adult Teratoma. J. Med. Imaging Health Inform. 2019, 9, 956\u0026ndash;961. [CrossRef] \u003c/li\u003e\n\u003cli\u003eEss, A.; M\u0026uuml;ller, T.; Grabner, H.; Van Gool, L. Segmentation-Based Urban Traffic Scene Understanding. BMVC 2009, 1, 2. \u003c/li\u003e\n\u003cli\u003eGeiger, A.; Lenz, P.; Urtasun, R. Are we ready for autonomous driving? The kitti vision benchmark suite. In Proceedings of the 2012 IEEE Conference on Computer Vision and Pattern Recognition, Providence, RI, USA, 16\u0026ndash;12 June 2012; pp. 3354\u0026ndash;3361. \u003c/li\u003e\n\u003cli\u003eMa, Z.; Tavares, J.M.R.S.; Jorge, R.M.N. A review on the current segmentation algorithms for medical images. In Proceedings of the 1st International Conference on Imaging Theory and Applications, Lisbon, Portugal, 5\u0026ndash;8 February 2009. \u003c/li\u003e\n\u003cli\u003eFerreira, A.; Gentil, F.; Tavares, J.M.R.S. Segmentation algorithms for ear image data towards biomechanical studies. Comput. Methods Biomech. Biomed. Eng. 2014, 17, 888\u0026ndash;904. [CrossRef] \u003c/li\u003e\n\u003cli\u003eMa, Z.; Tavares, J.M.R.S.; Jorge, R.N.; Mascarenhas, T. A review of algorithms for medical image segmentation and their applications to the female pelvic cavity. Comput. Methods Biomech. Biomed. Eng. 2010, 13, 235\u0026ndash;246. [CrossRef] \u003c/li\u003e\n\u003cli\u003eXu, A.; Wang, L.; Feng, S.; Qu, Y. Threshold-based level set method of image segmentation. In Proceedings of the Third International Conference on Intelligent Networks and Intelligent Systems, Shenyang, China, 1\u0026ndash;3 November 2010; pp. 703\u0026ndash;706.\u003c/li\u003e\n\u003cli\u003eCigla, C.; Alatan, A.A. Region-based image segmentation via graph cuts. In Proceedings of the 2008 15th IEEE International Conference on Image Processing, San Diego, CA, USA, 12\u0026ndash;15 October 2008; pp. 2272\u0026ndash;2275.\u003c/li\u003e\n\u003cli\u003eYu-Qian, Z.; Wei-Hua, G.; Zhen-Cheng, C.; Tang, J.-T.; Li, L.-Y. Medical images edge detection based on mathematical morphology. In Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, Shanghai, China, 17\u0026ndash;18 January 2006; pp. 6492\u0026ndash;6495.\u003c/li\u003e\n\u003cli\u003eHe, K.; Gkioxari, G.; Doll\u0026aacute;r, P.; Girschik, R. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 22\u0026ndash;29 October 2017; pp. 2961\u0026ndash;2969. \u003c/li\u003e\n\u003cli\u003eLin, G.; Milan, A.; Shen, C.; Reid, I. Refinenet: Multi-path refinement networks for high-resolution semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21\u0026ndash;26 July 2017; pp. 1925\u0026ndash;1934. \u003c/li\u003e\n\u003cli\u003eNoh, H.; Hong, S.; Han, B. Learning deconvolution network for semantic segmentation. In Proceedings of the IEEE International Conference on Computer Vision, Las Condes, Chile, 11\u0026ndash;18 December 2015; pp. 1520\u0026ndash;1528.\u003c/li\u003e\n\u003cli\u003eGu, J.; Wang, Z.; Kuen, J.; Ma, L.; Shshroudy, A.; Shuai, B.; Liu, I.; Wang, X.; Wang, G.; Cai, J.; et al. Recent advances in convolutional neural networks. Pattern Recognit. 2018, 77, 354\u0026ndash;377. [CrossRef] \u003c/li\u003e\n\u003cli\u003eHubel, D.H.; Wiesel, T.N. Receptive fields, binocular interaction and functional architecture in the cat\u0026rsquo;s visual cortex. J. Physiol. 1962, 160, 106. [CrossRef] [PubMed] \u003c/li\u003e\n\u003cli\u003eFukushima, K.; Miyake, S. Neocognitron: A self-organizing neural network model for a mechanism of visual pattern recognition. In Competition and Cooperation in Neural Nets; Springer: Berlin, Germany, 1982; pp. 267\u0026ndash;285. \u003c/li\u003e\n\u003cli\u003eL\u0026eacute;cun, Y.; Bottou, L.; Bengio, Y.; Haffner, P. Gradient-based learning applied to document recognition. IEEE 1998, 86, 2278\u0026ndash;2324. [CrossRef] \u003c/li\u003e\n\u003cli\u003eKrizhevsky, A.; Sutskever, I.; Hinton, G.E. Imagenet classification with deep convolutional neural networks. Adv. Neural Inf. Process. Syst. 2012, 60, 1097\u0026ndash;1105. [CrossRef] \u003c/li\u003e\n\u003cli\u003eHe, K.; Zhang, X.; Ren, S.; Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 16 June\u0026ndash;1 July 2016; pp. 770\u0026ndash;778. \u003c/li\u003e\n\u003cli\u003eSimonyan, K.; Zisserman, A. Very deep convolutional networks for large-scale image recognition. arXiv 2014, arXiv:1409.1556. \u003c/li\u003e\n\u003cli\u003eQiu, Z.; Yao, T.; Mei, T. Learning spatio-temporal representation with pseudo-3d residual networks. In Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 22\u0026ndash;29 October 2017; pp. 5533\u0026ndash;5541.\u003c/li\u003e\n\u003cli\u003eLong, J.; Shelhamer, E.; Darrell, T. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, 7\u0026ndash;12 June 2015; pp. 3431\u0026ndash;3440.\u003c/li\u003e\n\u003cli\u003eMyronenko, A. 3D MRI brain tumor segmentation using autoencoder regularization. In Proceedings of the International MICCAI Brainlesion Workshop, Shenzhen, China, 17 October 2018; pp. 311\u0026ndash;320.\u003c/li\u003e\n\u003cli\u003eNie, D.; Wang, L.; Adeli, E.; Lao, C.; Lin, W.; Shen, D. 3-D fully convolutional networks for multimodal isointense infant brain image segmentation. IEEE Trans. Cybern. 2019, 49, 1123\u0026ndash;1136. [CrossRef] [PubMed] \u003c/li\u003e\n\u003cli\u003eWang, S.; Yi, L.; Chen, Q.; Meng, Z.; Dong, H.; He, Z. Edge-aware Fully Convolutional Network with CRF-RNN Layer for Hippocampus Segmentation. In Proceedings of the 2019 IEEE 8th Joint International Information Technology and Artificial Intelligence Conference (ITAIC), Chongqing, China, 24\u0026ndash;26 May 2019; pp. 803\u0026ndash;806. \u003c/li\u003e\n\u003cli\u003eBorne, L.; Rivi\u0026egrave;re, D.; Mangin, J.F. Combining 3D U-Net and bottom-up geometric constraints for automatic cortical sulci recognition. In Proceedings of the International Conference on Medical Imaging with Deep Learning, London, UK, 8\u0026ndash;10 July 2019.\u003c/li\u003e\n\u003cli\u003eCasamitjana, A.; Cat\u0026agrave;, M.; S\u0026aacute;nchez, I.; Combalia, M.; Vilaplana, V. Cascaded V-Net using ROI masks for brain tumor segmentation. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 381\u0026ndash;391. \u003c/li\u003e\n\u003cli\u003eMoeskops, P.; Veta, M.; Lafarge, M.W.; Eppenhof, K.A.J.; Pluim, J.P.W. Adversarial training and dilated convolutions for brain MRI segmentation. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support; Springer: Cham, Switzerland, 2017; pp. 56\u0026ndash;64.\u003c/li\u003e\n\u003cli\u003eRezaei, M.; Harmuth, K.; Gierke, W.; Kellermeier, T.; Fischer, M.; Yang, H.; Meinel, C. A conditional adversarial network for semantic segmentation of brain tumor. In Proceedings of the International MICCAI Brainlesion Workshop, Quebec City, QC, Canada, 14 September 2017; pp. 241\u0026ndash;252. \u003c/li\u003e\n\u003cli\u003eGiacomello, E.; LoIacono, D.; Mainardi, L. Brain MRI Tumor Segmentation with Adversarial Networks. arXiv 2019, arXiv:1910.02717.\u003c/li\u003e\n\u003cli\u003eSimpson, A.L.; Antonelli, M.; Bakas, S.; Bilello, M.; Farahani, K.; Van Ginneken, B.; Kopp-Schneider, A.; Landman, B.A.; Litjens, G.; Menze, B.; et al. A large annotated medical image dataset for the development and evaluation of segmentation algorithms. arXiv 2019, arXiv:1902.09063\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"MRI Segmentation, Vision Transformers (ViT), Kernel-Based Convolution Neural Networks (CNN), and Multi-Class Support Vector Machines (M-SVM)","lastPublishedDoi":"10.21203/rs.3.rs-3377680/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3377680/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAccurate MRI segmentation is a crucial part of modern medical diagnostics and is essential for early disease diagnosis and effective treatment planning. Vision Transformers (ViT), Kernel-Based Convolutional Neural Networks (CNN), and Multi-Class Support Vector Machines (M-SVM) are all presented in this study as part of a novel hybrid approach to MRI segmentation that improves accuracy and efficiency.Our method employs ViT, which rapidly extracts high-level features from MRI patches, in combination with kernel-based convolutional neural networks, which are well-known for their ability to capture intricate patterns in image data. The M-SVM then refines the classification process, separating the pixels into distinct classes that are suggestive of different tissue types, and the segmentation phase begins without any problems. In addition to increasing the accuracy of MRI segmentation, initial findings suggest that this novel method might set an innovative standard for the analysis of medical images. This research has the potential to be an important development in medical imaging, which would significantly advance the current state of the art in healthcare technology by improving the accuracy with which diagnoses are made and the effectiveness of treatment plans.\u003c/p\u003e","manuscriptTitle":"Hybrid Approach for MRI Segmentation using Deep Learning and Machine Learning Algorithms","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2023-10-03 22:41:29","doi":"10.21203/rs.3.rs-3377680/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"7d44038d-4f94-4a45-ba31-aa109180c4bb","owner":[],"postedDate":"October 3rd, 2023","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2023-12-29T00:44:17+00:00","versionOfRecord":[],"versionCreatedAt":"2023-10-03 22:41:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3377680","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3377680","identity":"rs-3377680","version":["v1"]},"buildId":"7rjqhiLT3MXkJMwkYKINL","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}