2D Brain Tumor Classification Using CNN and SVM

The main objective is to devise a highly efficient and selfoperating brain tumor classification system characterized by elevated training and testing accuracy, a robust resistance to overfitting, and low computational complexity. Our model has successfully met all of these conditions. In our approach, Convolutional Neural Networks (CNNs) are employed for feature extraction, while Support Vector Machines (SVMs) are utilized for the classification task. By combining MRI and CT scan images, our model gains increased resilience. CNNs, specifically, excel in independently acquiring complex features from multimodal MRI images, capturing details from both normal brain tissues and tumor tissues for a comprehensive analysis. This methodology aims to contribute to the advancement of accurate and efficient brain tumor classification systems, ensuring optimal performance across various metrics. In our work, first we use CNN for feature extraction then we use SVM for classification in this model we are not using optimizer so the weights are not updated. Here we got the training accuracy 100% and testing accuracy 95.11%. The problem of this model is overfitting. Further for better result and remove the problem we use Adam optimizer for back propagation and updating the weight. Here we got the output training accuracy to 99.51%, validation accuracy 97.62 and testing accuracy 97.55. Further for better result we use RMSProp optimizer and we got the output training accuracy to 99.55%, validation accuracy 97.28 and testing accuracy 97.30. We required more accuracy so we further change some parameter, in previous model we are not using any padding so by apply convolution filter image size is reduced, in the previous model we used 32,64,128,256 convolution filter. In the final model we used 32, 64, 128, 256 and 512 convolution filter and all of the convolution filter we used zero padding then image size is not reduced. We employed the RMSProp optimizer for backpropagation, achieving significant success in training our model. With a dataset consisting of 2937 images for training and 327 images for testing, and employing a 20-epoch training procedure with a runtime of 285 milliseconds per step, we attained impressive results. Our model demonstrated exceptional performance with a accuracy in training is 99.55%, accuracy in validation is 98.30%, and accuracy in testing is 98.47%. Looking ahead, our future endeavors involve transitioning to 3D brain images, aiming to enhance the efficiency of brain tumor classification and segmentation. We aim to go beyond merely detecting tumors, striving to precisely outline their boundaries and forecast their dimensions for enhanced accuracy. Undoubtedly, delving into this ambitious endeavor will enhance our grasp of brain pathologies and elevate diagnostic capacities within the realm of medical imaging.

[1]               Hany Kasban, Mohsen El-bendary, Dina Salama, A comparative study of medical imaging techniques, Int. J. Inf. Sci. Intell. Syst. 4 (2015) 37–58; J. Clerk Maxwell, A Treatise on Electricity and Magnetism, vol. 2, 3rd ed., Clarendon, Oxford, 1892, pp. 68–73. [2]               D. Surya Prabha, J. Satheesh Kumar, Performance evaluation of image segmentation using objective methods, Indian J. Sci. Technol. 9 (8) (February 2016). [3]               brain-tumor : statistics, Cancer.Net Editorial Board, 03/2023 (Accessed on 24th Oct 2023) [4]               Ayadi W., Elhamzi W., Charfi I., Atri M. Deep CNN for Brain Tumor Classification. Neural Process. Lett. 2021;53:671–700. doi: 10.1007/s11063-020-10398-2. [CrossRef] [Google Scholar] [5]               Jayade S., Ingole D.T., Ingole M.D. Review of Brain Tumor Detection Concept using MRI Images; Proceedings of the 2019 International Conference on Innovative Trends and Advances in Engineering and Technology (ICITAET); Shegoaon, India. 27–28 December 2019; [CrossRef] [Google Scholar] [6] M.L.Oelze, J.F. Zachary, W.D. O’Brien Jr., Differentiation of tumour types in vivo by scatterer property estimates and parametric images using ultrasound backscatter, vol. 1, 5-8 Oct. 2003, pp. 1014–1017 [7]               Arkapravo Chattopadhyay, Mausumi Maitra, MRI-based brain tumour image detection using CNN based deep learning method in Neuroscience Informatics, feb 2022 www.elsevier.com/locate/neuri [8]          Tonmoy Hossain, Fairuz Shadmani Shishir, Mohsena Ashraf, M.D. Abdullah Al Nasim, Faisal Muhammad Shah, Brain tumour detection using convolutional neural network, in: 1st International Conference on Advances in Science, Engineering and Robotics Technology, ICASERT, 3-5 May 2019, 2019. [9] J. Seetha, S. Selvakumar Raja, Brain tumour classification using convolutional neural networks, Biomed. Pharmacol. J. 11 (2018) 1457–1461, https://doi.org/ 10.13005/bpj/1511 [10] MRI brain tumor detection using deep learning and machine learning approaches, Measurement: Sensors ,Volume 31, February 2024, 101026, https://doi.org/10.1016/j.measen.2024.101026