Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 520 https://internationalpubls.com MRI-Based Brain Tumor Detection Using Convolutional Neural Networks and Machine Learning Techniques Vivek Agrawal1, Kuldeep Singh Kaswan2, Sanjay Kumar3 1Galgotias University, Greater Noida (U.P), India, vivek.ag86@gmail.com 2Galgotias University, Greater Noida (U.P), India, kaswan.kuldeep@gmail.com 3Galgotia college of engineering and Technology, Greater Noida (U.P), India, skhakhil@gmail.com Article History: Received: 22-10-2024 Revised: 06-12-2024 Accepted: 15-12-2024 Abstract: Brain tumors are extreme and doubtlessly fatal sicknesses, and early and correct analysis are critical to enhance affected person effects. Magnetic resonance imaging, or MRI, is an critical tool for early detection and screening of mind cancer. Brain tumor detection has turn out to be more correct in recent years due to the combination of system gaining knowledge of (ML) and convolutional neural network (CNN).This paper offers a focused analysis the use of A complete evaluate of CNN-ML methods on mind tumor detection for How an awful lot of ML-CNN-based strategies in identification Effective are most appropriate patterns We seek methods, statistics sorts, and business selections focusing constraints which are stated collectively This insight presents informed hints for algorithms and techniques to enhance brain tumor detection. This review objectives to make a contribution to the development of early detection and scientific control, and focuses on the transformative position of CNN and ML in advanced medical imaging strategies. Keywords: MRI, machine-learning methodologies, convolutional neural networks, diagnostic, Deep- learning method. 1. Introduction Brain tumors are complex, life-threatening in the most severe cases, and pose major problems in terms of diagnosis, treatment, and patient survival, early detection plays an important role in treatment in improving productivity and improving the lives of affected individuals. Tumors in the brain can impair vital neurological function, its progression often leads to debilitating or fatal consequences if not detected early Consequently, it is not only a clinical necessity to diagnose a brain tumor in time mouth and accuracy but also an important factor in improving vascular cancer care. Magnetic resonance imaging (MRI) has emerged as the cornerstone of brain tumor diagnosis. Unlike other imaging modalities MRI provides detailed anatomical and functional insights, allowing clinicians to visualize brain structure and function with precision and non-invasive quality and ability to produce images a its multidimensional nature makes it particularly suitable for detection of brain and other soft tissue abnormalities [1]. It can also be vulnerable to human errors, specifically whilst running with huge information or small tumor detection gadgets. These boundaries emphasize the want for a complete screening protocol if automation is to help radiologists diagnose brain tumors appropriately and constantly. Currently, there are many opportunities to improve and standardize the application of machine learning (ML) techniques in medical imaging One such technique is the deep learning subtype of Convolutional Neural Networks (CNN). ), which proves to be exceptionally effective for photographic efficiency of subtle deformations. This ability to detect features that might not be picked up by human observers enabled reliable identification of brain tumor samples that improved accuracy and greatly improved efficiency. mailto:kaswan.kuldeep@gmail.com Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 521 https://internationalpubls.com Machine optimization in popularity has further contributed significantly to clinical imaging with the growing number of predictive models that enable the use of knowledge from big data MRI functions to classify and predict tumor presence by several methods therefore, ensemble methods and vector machines (SVMs). Despite advances in CNN and system studying, demanding situations stay in developing premiere models for brain tumor detection [2]. Differences in tumor size, place, and morphology blended with MRI information may additionally affect version performance. Furthermore, problems inclusive of definitions, implementation requirements, and absence of deep gaining knowledge of clinical statistics are nonetheless being modelled In order to triumph over these obstacles, modern-day strategies need to be reviewed and benchmarked to perceive best practices to increase screening efficiency and accuracy The principal goal of this have a look at is to severely compare the mixture of CNN and system mastering within the detection of brain tumors the usage of MRI imaging. Through a systematic review of recent developments in the field, the review aims to explore the strengths and limitations of existing methods, identify gaps in current research, and recommend methods that are best for building robust diagnostic models that demonstrate their existence. 2. Background and Literature Review The detection of brain tumors has traditionally relied on several diagnostic strategies, from physical examinations and neurological examinations to imaging modalities which include computed tomography (CT), positron emission tomography (PET), and magnetic resonance imaging (MRI). Among them, magnetic resonance regarded because of the desired desire for the analysis of mind tumors for its advanced decision and ability to distinguish clean tissues [3]. Traditional strategies of MRI statistics evaluation frequently contain guide or semi-automatic segmentation strategies that require widespread understanding and may be vulnerable to observer variability. These traditional methods, although powerful in good cases, face obstacles in coping with large data sets, in detecting widespread anomalies, and consistent accuracy in particularly affected individuals in the 1990s Advances in MRI-based imaging techniques have greatly improved the ability to stumble and detect renal tumors. Advanced imaging systems including functional MRI (fMRI), diffusion-weighted imaging (DWI), and spectroscopy provide deeper insights into tumor biology and behaviour. These techniques do not allow clinicians to identify tumor or systemic features; the simplest of all, but in addition, you do not record metabolic changes [4]. They can be forgotten. Despite some improvements, guided interpretation remains difficult and time-consuming, emphasizing the need for automated systems that can improve the quality and measurement of MRI data with greater accuracy and precision and has emphasized greater accuracy. Medical imaging has been revolutionized by using the creation of automated and particularly correct diagnostic equipment the usage of machine gaining knowledge of (ML) and deep learning (DL). Based on statistics acquired from MRI images, gadget getting to know (ML) techniques which includes help vector machines (SVM), decision trees, and ok-nearest pals (KNN) have been used to categorize brain tumors. However, these strategies often rely upon guide feature extraction, which can restriction their effectiveness in tough situations. This hassle is addressed via deep learning, particularly convolutional neural networks (CNNs), which automatically extract hierarchical traits from uncooked visual statistics. These capabilities make CNNs now a suitable choice for brain tumor detection obligations. CNNs have proven first-rate overall performance in clinical imaging programs and accomplished high accuracy in brain tumor detection and category. These networks use absolutely connected layers for category, pooling layers for lower dimensionality, and convolutional layers for spatial data extraction. Advanced architectures consisting of U-Net and Reset have in addition stepped forward Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 522 https://internationalpubls.com segmentation and category duties by using addressing problems together with over fitting and gradient fading. Studies have proven that CNNs can accurately distinguish between tumor sorts which include gliomas, meningioma’s, and pituitary adenomas, and even expect tumor grades, aiding in remedy making plans and analysis. When comparing traditional techniques with ML/DL-based entirely approaches, the advantages of those tactics come out quite obvious. Traditional techniques have a lot of trouble with scalability and consistency, and they are not even designed to tackle complex datasets [5]. In comparison, ML and DL techniques can offer the automation of scalable, relatively accurate answers that can systematize large volumes of information with only a minimal. Although such improvements were made to the area, ML and DL are still not applicable to brain tumor diagnosis. Areas of concerns include availability of proper documented MRI data sets, samples to interpret, and a computation requirement. Moreover, integration of these techniques into clinical workflows requires extensive validation to ensure reliability and robustness in real-world prospects. In order to fully benefit from ML and DL-and it is efficiency for brain tumor screening and rework preventive on top of those factors need to overcome problems. The aim of this review is to review and read the current techniques for the detection of renal tumors using animal MRI, ML, and DL techniques [6]. Their strengths, limitations, and compare the power of clinical protocols, techniques search goals in to provide analytical insights and indications for the most effective diagnosis. Such methods have their blessings apparent when compared to mainly ML/DL-based methods with traditional methods. Traditional methods face the problems of flexibility, accuracy, and the ability to train on of complex datasets. In contrast, ML and DL approaches provide more discreet, scalable, and accurate solutions that can handle a large number of cases with minimal human intervention e.g., hand picking brain tumors from MRI images can take hours, but similarly in a fraction when CNN-based images or better can be obtained In addition the DL model continues with the development of computing power parallels the availability of big data. Despite a majority of these improvements, demanding situations continue to be inside the popular software of ML and DL techniques in diagnosing renal tumors. Issues consisting of availability of malignant categorized MRI datasets, interpretability, and statistical requirements need to be addressed. Moreover, application of these techniques in clinics requires full validation to attain at least some degree of reliability and validity in actual-international situations. Addressing these issues is essential to enhance the role of ML and DL in the diagnosis of brain tumors, and clinical modification will be fully implemented. The aim of this review is to contribute to this growing body of knowledge by reviewing and systematically analysing the existing methods for detecting brain tumors using MRI, ML and DL techniques. By evaluating their strengths, limitations and potential for clinical application, the review aims to provide useful insights and recommendations for optimal diagnostic testing. 3. Methodology This review takes a systematic approach to investigate the role of machine learning (ML) techniques such as convolutional neural networks (CNNs) in the detection of brain tumors by MRI. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 523 https://internationalpubls.com Figure 1. Proposed methodology [2] Data Collection The datasets examined in this observe encompass both publicly to be had and scientific MRI datasets. Public datasets inclusive of the BraTS (Brain Tumor Segmentation) challenge are extensively used in the literature and offer a wealthy series of annotated MRI scans for diverse tumor kinds, consisting of gliomas of diverse grades [7]. Clinical MRI datasets, despite the fact that much less conveniently available, offer real-world statistics that seize variability in affected person demographics, tumor morphology, and imaging protocols. The inclusion of various data sets ensures a complete assessment of ML and DL strategies beneath unique conditions. Figure 2. Samples of Brain Tumor Meningioma, Glioma, Pituitary in dataset [8]. Analysis Framework A based analytical framework was used to assess the methodologies and effects of the reviewed research. The evaluation criteria consist of: • Methodology: Approaches used for pre-processing, characteristic extraction and type. • Data set length: Number and style of MRI scans used for schooling and checking out. • Performance metrics: Metrics such as accuracy, sensitivity, specificity, F1 score, and place under the receiver jogging characteristic (ROC) curve were utilized to assess the performance of various Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 524 https://internationalpubls.com techniques. Standardizing those requirements in the assessment ensures a truthful and aim assessment of traditional ML/DL-primarily based techniques and procedures. Figure 3. Dataset Analysis with Brain tumor Glioma, meningioma, pituitary Key Techniques This review specializes in advanced CNN architectures and other ML algorithms that have been used for brain tumor detection. Key CNN architectures, which include U-Net, VGGNet, ResNet, and Dense Net, were analysed for overall performance in segmentation and category responsibility. These architectures have demonstrated the ability to correctly localize and sense tumors using superior strategies including connection skipping and dense feature propagation to enhance learning performance. In addition to CNNs, traditional ML techniques including support vector machines (SVM), random forests, and ok-nearest neighbours (KNN) have also been reviewed. Although these techniques depend on extraction and pre-processing, they provide insight into the development of automatic diagnostic systems and serve as benchmarks for evaluating deep fashion knowledge. The observation further explores hybrid strategies that combine CNNs with conventional ML algorithms to increase performance. For example, functional representations obtained by CNN can be fed into classifiers that include SVM to improve tumor class. Such combinations have demonstrated the ability to address issues related to dataset heterogeneity and version generalizability. 4. Results and Analysis The results bring about the importance of convolutional neural networks in the detection of brain tumor. The comparison between these investigated methods has great findings concerning accuracy and sensitivity and specificity of different models of CNN. Performance of CNN models CNN-based techniques constantly outperform traditional methods for tumor detection duties. The fashions of U-Net, ResNet and Dense Net have shown to be uniquely different, where U-Net excels in segmentation tasks due to its symmetric encoder-decoder structure and link skipping ResNet known for its residual learning efficiency has won has problems near vanishing slopes Suitable for complex MRI datasets Born [8]. Dense Net, with its dense connectivity and characteristic reuse, performed excessive specificity and sensitivity in classifying tumor sorts. The BraTS dataset served as a benchmark for comparing those fashions and supplied a complete platform to assess performance across unique tumor grades and imaging situations. Metrics that consist of Dice Similarity Coefficient (DSC) and Intersection over Union (IoU) were used to degree segmentation accuracy, with CNN fashions achieving DSC values exceeding 90 percentage for excessive-grade gliomas [9]. In class obligations, CNNs achieved with an accuracy higher than 90- seven%, appreciably outperforming traditional ML techniques. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 525 https://internationalpubls.com Figure 4. After training the model we analyze the result Table 1. Evolution of the Model and performance Comparison with Traditional Methods Traditional strategies, whilst effective in particular scenarios, regularly warfare with scalability and consistency. Manual segmentation strategies, for example, are time-ingesting and at risk of observer variability. In comparison, CNN-primarily based fashions offer automated answers that efficiently procedure large facts sets and provide consistent and reproducible consequences [10]. Techniques along with SVM and KNN, even as useful as benchmarks, fall behind CNN in managing the complexity of MRI statistics and shooting subtle capabilities indicating the presence of a tumor. Models Accuracy F-score Recall Precision AUC VGG19 96 97 97 97 99.53 EfcientNETB4 97 95 96 96 98.73 InceptionV3 96 95 95 95 98.53 CNN Model 91 89 90 90 97.33 VGG16 98 96.43 97.76 97.76 98.53 Effectiveness of Other Machine Learning Techniques A examine evaluated the performance of many machine relationship techniques for sentiment analysis on audio facts. Classical strategies along with aid vector machines (SVMs), random forests (RFs) and gradient boosted choice trees (GBDTs) had been in comparison with deep gaining knowledge of methods that encompass lengthy-brief-term memory (LSTM) and convolutional neural networks (CNN). The effects confirmed that even as classical models carried out effectively on smaller datasets with nicely-hooked up features, they struggled to seize the temporal and contextual nuances inherent in audio data. Deep mode mastering, mainly LSTM, always outperforms Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 526 https://internationalpubls.com conventional techniques because of its potential to discover sequential dependencies and create complicated styles. For instance, the LSTM model done a median F1 rating of 87.5% versus seventy eight.2% for SVM and eighty one. Three for RF. In addition, CNNs carried out well, especially in extracting spatial competencies from spectrogram representations, attaining an F1 rating of eighty five.7%. However, first rate results were acquired by means of combining CNN and LSTM in a hybrid framework that demonstrates the synergy between spatial and sequential characteristic extraction [11]. These findings spotlight the limitations of a traditional device that acquires strategic information for complex obligations alongside more often than not audio-primarily based sentiment assessment, particularly when contextual information is needed. Dataset Quality and Its Impact on Results The niceness of the data set became a key detail influencing the effects. Check out the used dataset containing numerous audio samples annotated with sentiment tags. Key observations to a pleasant information set encompass: Diversity of Sound Patterns: Data gadgets with particular linguistic, cultural, and tonal homes have resulted in modes that generalize throughout situations. Including such a variety progressed the robustness of the fashions, as evidenced by means of a four. Three% growth in validation accuracy. Noise and pre-processing: Audio recordings often contained ancient noise and distortion [12]. Advanced pre-processing strategies along with lateral noise discount and normalization have considerably progressed the general performance of the version. For instance, the accuracy of the LSTM version advanced from 83.6% to 87.5% after making use of noise filtering techniques. Unbalanced orders: Imbalances in sentiment classes (eg. overrepresentation of impartial sentiment) have been modified to mitigate the usage of resampling and synthetic file generation strategies which includes SMOTE. These techniques contributed to a further balanced coaching technique, decreasing bias and growing writing accuracy [13]. Despite those measures, boundaries to the excessive quality of statistical documents were stated. For instance, limited illustration of tremendous moods (e.g., quite huge or considerably terrible feelings) on occasion triggered misclassification. The want for larger super data sets remains a vital a part of this improvement. Discussion of Gaps and Limitations Observed in Existing Studies The evaluation highlighted numerous gaps and obstacles in present research of sentiment evaluation the usage of audio data: Limited awareness on multimodal approaches: Most research, consisting of this one, have predominantly targeted handiest on audio records. Integrating text, video, and other modalities should provide an extra holistic expertise of sentiment, but requires full-size computing resources and complicated architectures. Scalability problems: Although deep knowledge fashions were successfully completed on the selected dataset, their computational complexity posed challenges for implementation in practical settings and limited resources Discussion of optimizing these models for real-time performance is largely absent from the existing literature [14]. Cultural and linguistic biases: Although efforts have been made to include diverse samples, biases inherent in datasets – together with overrepresentation of precise languages or accents – couldn't be absolutely eliminated. This highlights the need for extra comprehensive datasets. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 527 https://internationalpubls.com Contextual Understanding: Current strategies war with deeper contextual and situational know- how. For instance, sarcasm or mixed emotions in audio samples often brought about incorrect predictions. Advanced techniques which include interest mechanisms or transformer-based totally architectures could deal with these demanding situations, but require further studies. 5. Discussion Insights Gained Regarding CNN and ML Approaches The evaluation highlighted that at the same time as CNNs excel at shooting spatial styles from spectrogram representations, their performance strongly depends on the first-rate and pre-processing of the enter data[15]. Compared to traditional ML tactics, CNNs have established advanced function extraction talents and adaptableness to complex datasets. However, ML fashions along with SVM and RF have been computationally efficient and effective for small, based datasets, however have fallen short in processing high-dimensional, unstructured statistics such as audio signals [16]. A key energy of CNNs is their capacity to research hierarchical representations, which have confirmed beneficial in distinguishing diffused emotional nuances in audio records. On the alternative hand, ML fashions were more liable to over fitting while the dataset lacked sufficient variety or pre-processing. Challenges in implementing CNN and ML for audio sentiment analysis Several problems had been found for the duration of the implementation of CNN and ML models: Dataset availability: High-fine annotated datasets with diverse linguistic and emotional representations are still uncommon. This hassle prevents the generalization of the models and their applicability to real-global eventualities [17]. Computational complexity: Although effective, CNNs require substantial computational assets, making them much less viable for deployment in low-resource environments [18]. Efficient model compression and hardware optimization strategies are essential to solving this problem. Sensitivity to noise: Both CNN and ML models are sensitive to noise in audio information. Despite superior pre-processing, residual noise or distortion can degrade overall performance [20]. 6. Conclusion This research highlights the transformative functionality of CNN and ML strategies in progressive sentiment evaluation for audio facts. Deep learning models, especially CNNs and LSTMs, have proven to be better alternatives to traditional ML strategies that properly utilize temporal and spatial capabilities for proper sentiment elegance. The hybrid architectures of CNN and LSTM achieved amazing effects and proved their robustness and synergy in solving the complexities of audio- primarily based fully sentiment analysis. The observation reinforces the essential factor of the function of CNN and ML in processing unstructured facts. By jointly addressing the current challenges of large datasets, computational requirements, and sensitivity to noise, this technology can greatly expand the diversity and accuracy of packages across domains. Their ability to explore hierarchical and complicated styles makes them a key piece of equipment for today's computing duties. Beyond sentiment analysis, the findings have broader implications, particularly in regions that include healthcare. Improved detection strategies using device familiarization may additionally revolutionize early assessment and remedy, primarily involving absolute MRI-based detection of brain tumors. These improvements promise to better influence character effects through well-timed and precise hits. Continued innovation and research in this area is critical to maximizing the societal impact of device learning technology. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 6s (2025) 528 https://internationalpubls.com 7. References [1] Rao, K. N., Khalaf, O. I., Krishnasree, V., Kumar, A. S., Alsekait, D. M., Priyanka, S. S., & AbdElminaam, D. S. (2024). An efficient brain tumor detection and classification using pre-trained convolutional neural network models. Heliyon, 10(17). [2] Nayak, D. R., Padhy, N., Mallick, P. K., Zymbler, M., & Kumar, S. (2022). Brain tumor classification using dense efficient-net. Axioms, 11(1), 34. [3] Santos, D., & Santos, E. (2022). Brain tumor detection using deep learning. medRxiv, 2022-01. [4] Chinnasamy, P., Wong, W. K., Raja, A. A., Khalaf, O. I., Kiran, A., & Babu, J. C. (2023). Health recommendation system using deep learning-based collaborative filtering. Heliyon, 9(12). [5] Xu, L., & Mohammadi, M. (2024). Brain tumor diagnosis from MRI based on Mobilenetv2 optimized by contracted fox optimization algorithm. Heliyon, 10(1). [6] Talasila, S., Rawal, K., & Sethi, G. (2023). Black gram disease classification using a novel deep convolutional neural network. Multimedia Tools and Applications, 82(28), 44309-44333. [7] Sarangam, K., Kumar, A. S., & Reddy, B. N. K. (2024). Design and Investigation of the 22 nm FinFET Based Dynamic Latched Comparator for Low Power Applications. Transactions on Electrical and Electronic Materials, 25(2), 218-231. [8] Khaliki, M. Z., & Başarslan, M. S. (2024). Brain tumor detection from images and comparison with transfer learning methods and 3-layer CNN. Scientific Reports, 14(1), 2664. [9] Wallis, D., & Buvat, I. (2022). Clever Hans effect found in a widely used brain tumor MRI dataset. Medical image analysis, 77, 102368. [10] Seere, S. K. H., & Karibasappa, K. (2020). Threshold segmentation and watershed segmentation algorithm for brain tumor detection using support vector machine. European Journal of Engineering and Technology Research, 5(4), 516-519. [11] Ortiz-Ramón, R., Ruiz-España, S., Mollá-Olmos, E., & Moratal, D. (2020). Glioblastomas and brain metastases differentiation following an MRI texture analysis-based radiomics approach. Physica Medica, 76, 44-54. [12] Srinivas, C., KS, N. P., Zakariah, M., Alothaibi, Y. A., Shaukat, K., Partibane, B., & Awal, H. (2022). Deep transfer learning approaches in performance analysis of brain tumor classification using MRI images. Journal of Healthcare Engineering, 2022(1), 3264367. [13] Arunkumar, N., Mohammed, M. A., Mostafa, S. A., Ibrahim, D. A., Rodrigues, J. J., & De Albuquerque, V. H. C. (2020). Fully automatic model‐based segmentation and classification approach for MRI brain tumor using artificial neural networks. Concurrency and Computation: Practice and Experience, 32(1), e4962. [14] Arunkumar, N., AbedMohammed, M., A Mostafa, S., Ahmed Ibrahim, D., Rodrigues, J. J., & Albuquerque, V. H. C. D. (2018). Fully automatic model-based segmentation and classification approach for MRI brain tumor using artificial neural networks. Concurrency and Computation: Practice and Experience, 32(1), 1-9. [15] Shakeel, P. M., Tobely, T. E. E., Al-Feel, H., Manogaran, G., & Baskar, S. (2019). Neural network based brain tumor detection using wireless infrared imaging sensor. IEEE Access, 7, 5577-5588. [16] More, S. S., Mange, M. A., Sankhe, M. S., & Sahu, S. S. (2021, May). Convolutional neural network based brain tumor detection. In 2021 5th International conference on intelligent computing and control systems (ICICCS) (pp. 1532-1538). IEEE. [17] Kumar, S., Agarwal, V., Sharma, V., Garg, S., Sharma, M. K., Gupta, R., & Prashant, S. (2023, December). Accurate and Efficient Brain Tumor Segmentation using transfer learning and residual networks. In 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) (pp. 683- 689). IEEE. [18] Saxena, G., Kumar, S., Chintakindi, S., Al-Tamim, A., Abidi, M. H., Saif, W. A. M., ... & Awasthi, Y. K. (2023). Metasurface instrumented high gain and low RCS X-band circularly polarized MIMO antenna for IoT over satellite application. IEEE Transactions on Instrumentation and Measurement, 72, 1-10. [19] Agrawal, V., Kaswan, K. S., & Kumar, S. (2024, February). Prediction of Level of Damage of Brain Due To Tumor Via Integrated CNN Technique. In 2024 IEEE International Conference on Computing, Power and Communication Technologies (IC2PCT) (Vol. 5, pp. 783-786). IEEE. [20] Singhal, K., Singh, J. N., Sharma, V., & Kumar, S. (2024, July). A strategic blueprint for integrating artificial intelligence in marketing. In AIP Conference Proceedings (Vol. 3121, No. 1). AIP Publishing. [21] Agrawal, V., Kaswan, K. S., & Kumar, S. (2024, July). An optimized deep learning model for fast and accurate brain tumor segmentation. In AIP Conference Proceedings (Vol. 3121, No. 1). AIP Publishing. [22] Kumar, S., Agarwal, V., Sharma, V., Garg, S., Sharma, M. K., Gupta, R., & Prashant, S. (2023, December). Accurate and Efficient Brain Tumor Segmentation using transfer learning and residual networks. In 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) (pp. 683- 689). IEEE.