Abstract: This study conducted a comprehensive analysis of online brain tumor scan data, developing and evaluating a robust model to discern Intracranial Neoplasm brain tumors. Employing an empirical approach, we utilized Mask-RCNN and U-net for segmenting Intracranial Neoplasm tumors from brain Magnetic Resonance Images (MRIs). Our methodology encompassed dataset elucidation, data pre-processing, network architecture, training, testing strategies, and proposed ensemble methods. The dataset comprising 253 brain tumors and employed U-nets and Mask-RCNN, utilizing a statistical confusion matrix to evaluate tumor stratification performance, differentiating true positives (Tp), true negatives (Tn), false positives (Fp), and false negatives (Fn). Employing deep learning techniques and Python programming, using Accuracy, Sensitivity, Specificity, Dice coefficient, and Jaccard index metrics. Our findings revealed that the 176-training brain tumor dataset samples, 95 identified as "yes" tumors, while 5 misclassified as "no" tumors. For the testing set comprising 77 cases, 42 "yes" tumors were identified, along with 26 "no" tumors. The training model achieved remarkable performance metrics, boasting 93% accuracy, 95% sensitivity, and 90% specificity, with a 96% similarity to ground truth. The testing set results showcased 89% sensitivity, 86% Specificity, and 88% accuracy, across 253 cases, our model demonstrated 92% sensitivity, 88% specificity, and 90.5% accuracy. The model's runtime for segmenting the 176 datasets was 466 seconds, approximately 7 minutes and 45 seconds. In conclusion, our study yields highly satisfactory results with an accuracy of 90.5%, 92% sensitivity, and 88% specificity. This model exhibits promising potential for precise brain tumor boundary detection, enhancing segmentation, diagnosis, and guiding brain tumor surgery.

Keywords: Machine Learning; Convolutional Neural Network; U-nets, Mask-RCNN; Segmentation; Magnetic Resonance Images; and Brain Tumor.

Title: A Deep Learning Hybridized Model for Segmentation of Medical Brain Tumors

Author: Gani Timothy Abe, Philemon Uten Emmoh

International Journal of Novel Research in Computer Science and Software Engineering

ISSN 2394-7314

Vol. 11, Issue 2, May 2024 - August 2024

Page No: 25-44

Novelty Journals

Website: www.noveltyjournals.com

Published Date: 11-June-2024

Abstract: In the realm of mobile edge computing, enhancing Quality of Service (QoS) and security remains a pivotal challenge. Traditional approaches often struggle to balance computational efficiency with evolving network demands. To address these limitations, this study introduces a novel multimodal bio-inspired optimization framework, leveraging the strengths of Grey Wolf Optimizer (GWO), Firefly Optimizer (FFO), and Bat Algorithm process. The Grey Wolf Optimizer, selected for its proficiency in optimizing continuous non-linear functions, is adept at resource allocation, load balancing, and network routing within dynamic environments for different scenarios. This characteristic is crucial for mobile edge computing, where adaptability and responsiveness to fluctuating demands are key. Similarly, the Firefly Optimizer, known for its efficacy in large search spaces, is utilized for optimizing network configuration, security policy management, and energy efficiency. Its ability to evolve network configurations and security policies over time is indispensable in countering emerging threats and adapting to changing conditions. The Bat Algorithm, specializing in spatial optimization, is employed for server placement, enhancing energy efficiency, and reducing service latency. This is particularly vital in edge computing networks, where strategic server placement can significantly impact overall network performance levels. The implementation of these bio-inspired algorithms demonstrates substantial improvements over existing methods. When tested on UCI Datasets and IEEE Data Port Datasets, the proposed framework showed a 4.9% better makespan, a 5.5% higher deadline hit ratio, a 3.5% increase in cloud efficiency, and an 8.5% reduction in delay levels. These enhancements not only signify the efficacy of the proposed model in real-world scenarios but also highlight its potential in revolutionizing mobile edge computing. This work, therefore, represents a significant leap forward in optimizing edge computing networks, offering a robust, adaptable, and efficient solution for the ever-evolving demands of network infrastructure scenarios.

Keywords: Mobile Edge Computing, Bio-Inspired Optimization, Quality of Service, Network Security, Resource Allocations.

Title: Design of an Efficient Model for Enhancing Quality of Service and Security in Mobile Edge Computing through Multimodal Bio-Inspired Optimizations

Author: Dr. Jaya Manoj Gadge

International Journal of Novel Research in Computer Science and Software Engineering

ISSN 2394-7314

Vol. 11, Issue 2, May 2024 - August 2024

Page No: 11-24

Novelty Journals

Website: www.noveltyjournals.com

Published Date: 05-June-2024

Abstract: Brain tumor detection plays a crucial role in healthcare, significantly impacting patient survival rates. Brain cancer is one of the most severe and life-threatening conditions, necessitating early and accurate detection for effective treatment. Unfortunately, many patients in remote areas cannot access advanced diagnostic tools and treatment options, exacerbating the severity of the disease. Timely detection of brain tumors is essential to improve prognosis and survival rates. In this study, we explore the application of machine learning, particularly deep learning techniques, in detecting brain tumors from medical images. Our research focuses on the comparative analysis between convolutional neural networks (CNN) and the AlexNet architecture for brain tumor detection. The study demonstrates that CNN algorithms are highly effective, achieving an accuracy of 96.15%, while the AlexNet architecture, a specific type of CNN, achieved an even higher accuracy of 98.03%. The results indicate that AlexNet outperforms the standard CNN in terms of accuracy. Additionally, an Android application has been developed to facilitate brain tumor detection. This mobile app allows users to upload medical images, which the app analyzes to predict the presence of a tumor and provides relevant information to guide further medical consultation.

Keywords: Brain Tumor Detection, Image Preprocessing, AlexNet, Convolutional Neural Network, Deep Learning, Web App.

Title: DETECTION AND LOCALIZATION OF BRAIN TUMOR USING DEEP LEARNING AND COMPUTER VISION

Author: Dr. A.D. Katre, Aman Pathan, Reva Bhagwat, Gourav Kumbhar, Atharwa Wakade, Amra Nadaf

International Journal of Novel Research in Computer Science and Software Engineering

ISSN 2394-7314

Vol. 11, Issue 2, May 2024 - August 2024

Page No: 1-10

Novelty Journals

Website: www.noveltyjournals.com

Published Date: 29-May-2024