Cloud Enabled Enhanced AIML Based Diagnosis of Brain Tumors from MRI Data

Abstract

Abstract newlineThe brain is one of the most vital and delicate organs of the human body. It contains highly newlinespecialized cells responsible for controlling and coordinating the entire body. However, due newlineto uncontrolled cell growth, a brain tumor may develop, leading to severe complications and newlineeven life-threatening conditions. Early detection plays an important role in improving the newlinetreatment process and increasing the chances of recovery. Radiologists mostly rely on Magnetic newlineResonance Imaging (MRI) to detect brain tumors, as MRI provides excellent tissue contrast newlineand detailed imaging. Still, analysing MRI scans manually is time-consuming and requires newlinespecialized clinical expertise. Variations in individual interpretation among radiologists may newlinesometimes delay diagnosis and impact timely treatment decisions. newlineTo overcome these challenges, this research presents an intelligent Computer-Aided newlineDiagnosis (CAD) system capable of automatically detecting and classifying brain tumors from newlineMRI. The study progresses through three major development phases, combining the strengths newlineof machine learning (ML), deep learning (DL), and cloud-based automation to create a highly newlineaccurate, fast, and scalable diagnostic framework. newlineIn the ML-based phase, a hybrid segmentation and classification framework is introduced. newlineRough C-Means clustering is used to extract tumor regions effectively, while Grey Wolf newlineOptimization (GWO) improves segmentation precision by refining cluster parameters. The newlinesegmented output is analysed using a Radial Basis Function Gaussian Neural Network newline(RBFGNN) classifier, resulting in a classification accuracy of 99.29 percent. This hybrid ML newlinedesign provides a strong foundation and establishes reliable tumor and non-tumor identification. newlineIn the DL-based phase, deeper and more powerful architectures are explored to enhance newlineboth segmentation and classification. Swin U-Net Transformer is used to capture long-range newlinedependencies and multi-level features in MRI scans, while EfficientNet V2 is adopted newlinefor classification due to its balanced structural efficiency and high accuracy. Using newlinethe BRATS 2020 dataset, the model achieved 78 percent segmentation accuracy and 97 newlinepercent classification accuracy. Another DL architecture combining U-Net for segmentation newlineand ResNet-50 for multi-class tumor recognition further improved performance, accurately newlineclassifying glioma, meningioma, and pituitary tumors with an accuracy of 99.17 percent. newlineIn the final cloud-enabled phase, the system is extended into a real-time diagnostic platform newlinethat hospitals and diagnostic centres can integrate into their workflows. The cloud deployment newlinesupports secure MRI upload, automated processing, and instant result generation. Continuous newlineIntegration and Continuous Deployment (CI/CD) pipelines enable automated model retraining, newlinevii newlinevalidation, and version updates, ensuring that the system remains accurate and up-to-date. newlineReal-time analysis, multi-user support, and efficient cloud storage make the system accessible newlineeven in remote healthcare locations. Evaluation using accuracy, precision, recall, F1-score, newlineand Jaccard Index confirms that the cloud-enabled CAD system outperforms many traditional newlinediagnostic methods. newlineOverall, by combining the strengths of ML, DL, and cloud-based CI/CD automation, this newlineresearch contributes a powerful and clinically relevant diagnostic solution that supports early newlinebrain tumor detection, improves treatment planning, and enhances patient outcomes through newlinefast, reliable, and consistent analysis. newlineKeywords: newlineBrain Tumor, Classification, Segmentation, Machine Learning, Deep Learning, ResNet, newlineUNet, Swin-UNet, EfficientNet, Grey Wolf Optimization (GWO), Cloud Computing, CI/CD newlinePipeline, Internet of Medical Things (IoMT), MRI Image Processing, Medical Image newlineAnalysis, Feature Extraction, Intelligent Healthcare Systems. newline