MRI-Based Brain Tumor Detection and Diagnosis

MRI-Based Brain Tumor Detection and Diagnosis

Mrs. N.Asha Latha

Assistant Professor, Department of CSE, Annamacharya Institute of Technology, and Sciences, Tirupati-517520, A.P.

Nageswari G

UG Scholar, Department of CSE Annamacharya Institute of Technology, and Sciences, Tirupati-517520, A.P.

Nikhila K

UG Scholar, Department of CSE Annamacharya Institute of Technology, and Sciences, Tirupati-517520, A.P.

Sravani B

UG Scholar, Department of CSE, Annamacharya Institute of Technology and Sciences, Tirupati-517520, A.P.

Manjunatha A

UG Scholar, Department of CSE, Annamacharya Institute of Technology and Sciences, Tirupati-517520, A.P.

Abstract- Brain tumor detection through the assistance of Magnetic Resonance Imaging (MRI) plays an important role in the diagnosis and treatment of the multi-systemic disorder. It is one of the medical image analysis techniques that detect and locate a tumor in one of the most important and sensitive organs: the human brain. This paper proposes an approach to develop an efficient and more accurate deep learning and machine learning based approach to detect and locate a tumor through images captured by an MRI. By using a Convolutional Neural Network (CNN), one of the intelligent deep learning approaches, features of an image and areas of tumors within an image will be automatically extracted. and Support Vecctor Machine(SVM), Is used to detect the tumor.Accuracy, precision, recall, and F1-score methods were used to evaluate this approach. Similarly, a graphical interface was added to see images.

Keywords: Brain Tumor Detection, MRI, CNN, Deep Learning, Image Segmentation, Medical Image Processing.

1. INTRODUCTION

   Brain tumors are among the neurological diseases that are very serious, and early detection plays a significant role in effective treatment an patient survival. MRI is widely used for diagnosis of brain tumors due to its better contrastive ability and its detailed images of brain tissues. But the manual examination of MRI images often requires much time and largely depends on expert radiologists, which may further delay or result in inconsistent outcomes.

   The proposed system overcomes the limitations by applying a Deep Learning-based automatic brain tumor detection system. This work applies CNNs because CNNs have powerful feature extraction capabilities from medical images. MRI image preprocessing is done to improve image quality and accuracy for the detection by resizing, normalizing, removing noise, and extracting the tumor region.

   The SVM model trained will detect whether a brain tumor exists, and classify the MRI input image. During prediction, some more image processing techniques will be done in order to localize the tumour region. The automatic approach

   decreases human effort and increases accuracy while supporting reliable analysis of medical images.

   Fig 1: Sample MRI Brain Image with Tumor

2. LITERATURE SURVEY

   Detection of a brain tumor from MRI images has gained major importance because early diagnosis improves treatment planning and the survival rate. MRI is widely preferred due to its high resolution and strong tissue contrast, which supports identification of abnormal growth in the brain. MRI images may contain noise and intensity variations that reduce clarity and affect the accurate interpretation of images. Hence, medical image preprocessing and enhancement techniques are generally used to improve image quality before applying any kind of classification and segmentation models [2224].

   The common concept behind traditional brain tumor detection systems is based on image processing and handcrafted feature extraction methods[26],[27]. Several works have utilized segmentation and filtering techniques, which include watershed-based approaches to segment the tumor region from MRI images, and objective measures for the assessment of segmentation results[27],[28]. Feature extraction methods such as texture and shape descriptors, among others, have been utilized with machine learning classifiers such as Support Vector Machine (SVM) for the identification of tumors. Reviews for MRI brain tumor segmentation have underlined the importance of accurate

   extraction of tumor boundaries and also discussed a few major obstacles, which are edema, necrosis, and heterogeneity in tumors[25].

   In the last few years, deep learning methods have demonstrated various advantages over traditional ML approaches[19],[23]. Convolutional Neural Networks (CNNs) automatically learn discriminative features from MRI images, which significantly improve classification performance compared to classical feature engineering- based methods. Various research has proved that the CNN models based on transfer learning improve tumor classification with reduced training data and better generalization. This approach improves classification accuracy and reduces overfitting issues and was addressed using DenseNet-based transfer learning methods. Recently, due to higher accuracy with optimized parameter usage, EfficientNet architectures have gained a lot of attention, and several works proposed improved EfficientNet models for multi-grade brain tumor detection, EfficientNet-B9-based high precision classification, and EfficientNet-B0-based multi-class detection systems[5],[6],[9].Lightweight CNN models have also been proposed, which, with limited training data, can achieve highly accurate detection of brain tumors and therefore are suitable for low-resource environments [4].

   Apart from classification, tumor segmentation plays a critical role in diagnosis because it shows the exact location and shape of the tumor.U-Net based segmentation models are widely used for segmenting tumor regions from MRI images. Enhanced variants of the U-Net have been introduced for improved performance in tumor segmentation tasks [15]. Enhanced variants such as multi-scale attention U-Net with EfficientNet as an encoder show improvement in MRI segmentation with better use of detailed contextual information for image segmentation tasks [11]. Researchers have introduced novel U-Net variants for image segmentation and better image segmentation architectures such as SLCA-UNet for brain tumor MRI image segmentation tasks [14].These segmentation methods contribute to better visualization and clinical decision support by highlighting the tumor-affected area clearly.

   For this reason, more hybrid and ensemble methods have been implemented to enhance reliability as well. Ensemble models of CNN-based deep learning prediction have shown more reliability and accuracy than relying on a singular prediction model [10]. Furthermore, SVM prediction models are still being explored in terms of tumor prognosis and classification using optimized feature selection algorithms, which indicates that SVM has potential application in tumor diagnostics [18]. Even the combination of One Class SVM and CNN has demonstrated potential in improving tumor detection in MRI scans [17]. Innovation in SVM using innovative feature extraction algorithms has shown potential in improving multiclass tumor detection [16].

   Some of the recent research trends include the applications of the Transformer architecture and the usage of the pretrained models in the classifications of brain tumors[3],[23]. From the literature, it was established that the pretrained models and the transformer architecture can aid in achieving higher accuracy in the classifications of the tumors in the brains of the human body as well as in the diagnosis of the same conditions[3],[20]. Impoved deep learning models were proposed for the detection of the tumor in the brains of the human body in order to improve the detection of the same[2],[21].

3. EXISTING SYSTEM

   The existing method has been discussed, which develops a system that uses machine learning, neural networks, deep learning, computer vision, etc., to spot brain tumor images from MRI scan images. MRI scan images are chosen as they have high resolution, good tissue contrast, etc., which are significant in detecting abnormal tissues in the brain region, especially brain tumor detection.

   The images will contain nearly 5000 images, both tumor and non-tumor images, of MRI brain images. At first, preprocessing will be done as its a crucial step in this project, followed by cleaning the images, where MRI images are

   cleaned to attain a clear skull-image, and unclear images are removed from the database. Then, images are sized to a consistent measure, i.e., images are resized to a specified number of dimensions, which helps in proper classification.

   Then, the images are split into training sets and testing sets, where classification algorithms, such as CNN and SVM, are implemented to classify tumor images from MRI scan images. Then, accuracy, precision, etc., are calculated, comparing both classifiers, where in the existing system, a CNN-based system has a high accuracy of 93%, while an SVM-based system has a lower accuracy of 83%.

   Fig 2: Existing System Architecture Using CNN and SVM

4. PROPOSED SYSTEM

   Fig 3: Proposed CNN & SVM-Based Brain Tumor Detection and Classification System

   The proposed system aims to develop a brain tumor detection model by employing an automated method through Deep Learning technologies to process images obtained through MRI scans. For this purpose, a Convolutional Neural Network (CNN) architecture will be implemented using labeled brain images obtained through an MRI scan that contain both tumor and non-tumor segments. Firstly, images are obtained from an MRI scan, resized, changed to grayscale, normalized, as well as data augmented to enhance the models generalization capabilities. In the proposed approach, the model has multiple convolutional network layers combined with multiple dense network layers with the use of “batch normalization” to obtain more accurate “high- level features.” In the proposed model, the model includes multiple “convolution network layer”s combined with multiple dense network layers through “batch normalization” to obtain more accurate “high-level features.”

   In the proposed approach, the model uses the “Adam” optimization technique. In the case where we want to predict the tumors, the model trained will be loaded to predict the tumoral regions by being applied to the MRI images. In image preprocessing for the prediction model, we remove the noise from the images, use the threshold operation to separate the regions of interest from the background.This processed image is then resized and normalized before being passed to the trained CNN model for prediction. Then, a system output is generated, indicating the presence or absence of a brain tumor. Furthermore, a graphical user interface is implemented into the system to provide a better visualization of input images.It will also provide an efficient, effective, as well as automated solution for early detection of brain tumors with fewer dependencies.,

5. RESULTS

   The proposed Brain Tumor Detection System has been successfully developed using a Convolutional Neural Network (CNN) model for the analysis of MRI brain images for the detection and classification of brain tumors. The proposed webbased system supports the upload of MRI

   images in PNG or JPG format for automatic tumor prediction.

   As depicted in Fig 4, the home page of the proposed system provides the user with the facility to upload the MRI scan through the drag-and-drop feature or file browser. After selecting the image, the user can start the analysis process by clicking on the Analyze MRI button.

   Once the MRI image is uploaded, the system analyzes the image and shows the result of the prediction outcome in the Prediction Result panel, as shown in Fig 5. The model provides the type of tumor identified and the confidence level of the result.

   Fig 5 shows a sample prediction result obtained from the system. In this case, the system was able to predict the MRI image as Meningioma with a confidence level of 0.63 and the tumor severity as Low Severity. This indicates that the proposed system is capable of providing accurate results for tumor identification.

   Once the MRI image is uploaded, the system analyzes the image and shows the result of the prediction outcome in the Prediction Result panel, as shown in Fig 5. The model provides the type of tumor identified and the confidence level of the result.

   Fig 5 shows a sample prediction result obtained from the system. In this case, the system was able to predict the MRI image as Meningioma with a confidence level of 0.63 and the tumor severity as Low Severity. This indicates that the proposed system is capable of providing accurate results for tumor identification.

   Model	Accuracy	Precision	Recall	F1- Score
   CNN+SVM	98.93%	98.95%	98.93%	98.93%

   Performance of Proposed CNN-SVM Model

   Fig 4: Brain Tumor Detection System Home and MRI Upload Interface

   Fig 5: Sample MRI Image and Tumor Prediction Result (Meningioma Detection)

   Fig 6: Brain Tumor Classification Report Showing Precision, Recall, F1-Score and Accuracy

6. CONCLUSION

   In this project, an automated system for detecting brain tumors using deep learning on MRI brain images was developed. The proposed system extracts relevant features from MRI images and assists in detecting the presence of brain tumors with minimal human intervention. Image preprocessing and additional processing steps help improve detection performance and support tumor localization. The systems performance is evaluated using standard metrics, demonstrating its potential effectiveness for automated brain tumor detection. additional processing steps help improve detection performance and support tumor localization. The systems performance is evaluated using standard metrics, demonstrating its potential effectiveness for automated brain tumor detection.

7. REFERENCES

1. Uniyal, Manas, et al. “Automated brain tumor detection using advanced deep learning models.” Discover Artificial Intelligence (2026).
2. Gupta, Siddhi, and Azrina Abd Aziz. “Advancing Brain Tumor Classification through Pre-Trained Transformer and Transfer Learning Models.” Franklin Open (2026).
3. Naeem, Awad Bin, et al. “Lightweight CNN for accurate brain tumor detection from MRI with limited training data.” Frontiers in Medicine 12 (2025).
4. Ishaq, Ahmad, et al. “Improved EfficientNet architecture for multi- grade brain tumor detection.” Electronics 14.4 (2025).
5. Muthulakshmi K and M. Jayalakshmi. “Rethinking model of EfficientNet-B9 for Brain Tumor Classification

   A High-Precision Deep Learning Approach.” Results in Engineering

   (2025).

6. Rasheed, Mehwish, et al. “Improved brain tumor classification through DenseNet121 based transfer learning.” Discover Oncology

   16.1 (2025).

7. Ahmad, Jameel, et al. “LIU-NET: lightweight inception U-Net for efficient brain tumor segmentation from mutimodal 3D MRI images.” PeerJ Computer Science 11 (2025).
8. Naseer, Mohammed Muddasir, et al. “An Optimized EfficientNet-B0 Framework for Multi-Class Brain Tumour Detection and Classification from MRI Images.” Biomedical and Pharmacology Journal 18.October Spl Edition (2025).
9. Eleyan, Gülden, et al. “Brain Tumor Detection Via Ensemble CNN- Based Deep Learning Models.” 2025 6th International Conference on Bio-engineering for Smart Technologies (BioSMART). IEEE, 2025.
10. J S Nisha, et al “Brain tumor segmentation using multi-scale attention U-Net with EfficientNetB4 encoder for enhanced MRI analysis.” Scientific Reports 15.1 (2025).
11. Pourmahboubi, Amin, Nazanin Arsalani Saeed, and Hamed Tabrizchi. “A brain tumor segmentation enhancement in MRI images using U-Net and transfer learning.” BMC Medical Imaging 25.1 (2025).
12. Obayya, Marwa, et al. “A novel U-net model for brain tumor segmentation from MRI images.” Alexandria Engineering Journal 126 (2025).
13. Tejashwini, P. S., J. Thriveni, and K. R. Venugopal. “A novel SLCA- UNet architecture for automatic MRI brain tumor segmentation.” Biomedical Signal Processing and Control 100 (2025).
14. Ananth, R. Prem, et al. “MRI Based Brain Tumor Segmentation with Enhanced U-Net Deep Learning Model.” 2025 3rd ICDSNS. IEEE, (2025).
15. Mjahad, Azeddine, and Alfredo Rosado-Muñoz.

    “Optimizing Tumor Detection in Brain MRI with OneClass SVM and Convolutional Neural Network-Based Feature Extraction.” Journal of Imaging 11.7 (2025).

16. PUNITHA V , R. THILLAIKARASI. “BRAIN TUMOR PROGNOSIS USING SUPPORT VECTOR MACHINE WITH

    OPTIMAL FEATURE SELECTION.” Journal of Tianjin University Science and Technology,(2025).

17. Beaulah Princiba D , P. Ezhilarasi, and S. Rajeshkannan. “Integrated EfficientNetB3V2 fused MaxEnt classifier model for brain tumor classification in MR images.” Journal of the Chinese Institute of Engineers 48.3 (2025).
18. G. Sudheer , Ganesan V “Detecting and Diagnosing Brain Tumor MRI Image,” IITCEE(2024).
19. Basthikodi, Mustafa, et al. “Enhancing multiclass brain tumor diagnosis using SVM and innovative feature extraction techniques.” Scientific Reports 14.1 (2024).
20. Rasa, Sadia Maduri, et al. “Brain tumor classification using fine- tuned transfer learning models on magnetic resonance imaging (MRI) images.” Digital health 10 (2024).
21. Deepak S , P M Ameer. “Brain tumor classification using deep CNN features via transfer learning.” Computers in biology and medicine 111 (2019).
22. Litjens, Geert, et al. “A survey on deep learning in medical image analysis.” Medical image analysis 42 (2017).
23. Telrandhe, Swapnil R., Amit Pimpalkar, and Ankita Kendhe. “Detection of brain tumor from MRI images by using segmentation & SVM.” 2016 world conference on futuristic trends in research and innovation for social welfare (Startup Conclave). IEEE, 2016.
24. Prabha, D. Surya, and J. Satheesh Kumar. “Performance evaluation of image segmentation using objective methods.” Indian J. Sci. Technol 9.8 (2016).
25. Kasban, Hany, M. A. M. El-Bendary, and D. H. Salama. “A comparative study of medical imaging techniques.” International Journal of Information Science and Intelligent System 4.2 (2015).
26. Shanthakumar, P., and P. Ganesh Kumar. “Computer aided brain tumor detection system using watershed segmentation techniques.” International Journal of Imaging Systems and Technology 25.4 (2015).
27. Gordillo Castillo, Nelly, Eduard Montseny Masip, and Pilar Sobrevilla Frisón. “State of the art survey on MRI brain tumor segmentation.” Magnetic resonance imaging (2013).
28. Cha, S. “Update on brain tumor imaging: from anatomy to physiology.” American Journal of Neuroradiology 27.3 (2006).